Home

Performance Manager

1. 2 sess 279 Performance Evaluation Report 2 ssa 280 Process Statistics aNG e e re Rer PERGERET iaaa Aan En aE aai Ea 282 Tabular Report Sections 2 tenet nnn 284 Graphs dassdasa dade RE RR RRERAPRI RR AEEA EE E RR ERR RRRERPPARRE RR 286 How YOU CUSTOMIZE 2 26 oth ye hte esce etes ette etiaro RG bka 296 Customize the Data Collection sse 296 Customize the PSDCSDATABASE Definition s 305 Customize Parameters eh 306 Workload Definitions sare 308 Workload Family Definitions ss rre 313 History File Descriptors sss 316 Parameter Settings 2 2 2 2 6 ccc ce lulii llis eder e a nr rena 319 View the Main Window 2 222 srl 320 Chapter 9 Use the DECwindows Motif Real time Display 321 Start the Real time Display 2 2 22 sea 321 Control the Real time Display ssssslsslsllslls sese 322 Navigate Within the Default Panels 22 2 sess 323 Use the Panel Commands Menu susssssssslsllss s e ns 324 Default Panel Descriptions s els 324 System OVervieW a na seca er re RR Rer e ECERRSFPECEPRCETE PGEdEG ED EdD EG ED Sad ELE 324 Default Panel Hierarchy 222 a 325 CPU Utilization Panel Descriptions een ence ene n ees 325 CPU Queue Panel Descriptions osses 3 aaa AG NG KA s 329 Hard Fault Rate Panel Descriptions e 330 Disk Rate Panel Descriptions sss 335 Review Data in Playback Mode
2. 2 sea 374 Display Multi node Statistics sss ls 376 Display Single Node Statistics srl 378 Display CPU Utilization 1 7015 aa bee PIG I LIIT DING BIDII DD PADDNG DAA vepRPPPPPPG Y ep 379 Display Top Processes Statistics AA teens 380 Display Top Device Statistics re 381 Display Process Information sss 381 Display Disk Information esses 384 Display Rules Information sr ln 385 Display RESOURCE Information s ls 385 RESOURCE Keypad eene dae E pae qr p E pPEVQPG dag iq dea daga eques 386 Balance Cluster System Utilization Using the Resource Display 2 22 005 386 Lower Common Resource Display 22222 ses 387 Memory Displayt M 388 Disk Display o ovest pl e EE p ba e pakay maba bu ple ble poe bukal Ive C us 389 CPU Display iiid a usc aca Xa add da da i A IG NG qudd qud qaa r4 eee dees a i3 390 The INVESTIGATE Command 2 2222 rr 391 INVESTIGATE Command Options 2 2 22 391 Contents 9 INVESTIGATE Keypad 0 00 ccc he 392 Evaluate Performance Using the Investigate Displays 0 002 c cece eee eee 392 Investigate a Memory EIMIta tiom seriernes erica tra eine need ease eed nated eae nee 394 Investigate an I O Limitation nananana anaana e eens 395 Investigate a CPU Limitation 0 cent en raaraa 396 Isolate the Cause of a Memory Lim
3. Hoocmn Bomncocoocou cOoOoocococcococo DD DDQD QD QD QO The following statements are keyed to the columns in the previous example 1 Node name 2 Tape controller 3 Unit number 4 Percentage of records having tape activity on a given node during the interval UI 1 Os per second to the tape controller unit 6 Errors per second 7 Tape controller unit type 60 Performance Manager Administrator Guide Performance Evaluation Report Interpret cluster wide Hot File Statistics Performance Evaluation D 6 Ops Ops Time Dev Avg Peak Rds Spl DD BORDEAUX 3 DUA29 1 ZINFANDEL 3 DUA30 3 25 3 42 55 0 0 03 0 03 40 O 2 0 03 100 50 2 0 02 100 O 0 0 0 0 In the Performance Evaluation Report the hot file statistics follow the cluster wide tape statistics The hot file statistics highlight the files with the most I O operations for each disk in the configuration By default the 20 hottest files are provided for each disk however you can change this number by specifying HOTFILE LIMIT n on the command line To display only the hot file statistics section of the report use the following qualifier INCLUDE HOTFILE STATISTICS To disable the hot file statistics display from the Performance Evaluation Report use the following qualifier INCLUDE NOHOTFILE STATISTICS The following example shows the hot file statistics section of the Performance Evaluation report CLUSTER Page 1 PA Vx x Wedne
4. 184 Performance Manager Administrator Guide ADVISE PERFORMANCE EXPORT The following table lists the keywords Keyword USERNAMES string IMAGENAMES string PROCESSNAMES string ACCOUNTNAMES string WORKLOADNAMES workloadname Description Specify FILTER USERNAMES to export all process records with the username matching any of the specified strings Specify FILTER IMAGENAMES to export all process records with the imagename matching any of the specified strings Do not specify any trailing EXE nor the file version device or directory Specify FILTER PROCESSNAMES to export all process records with the processname matching any of the specified strings The match string is case sensitive so if the process names have any lowercase letters spaces or tabs use double quotes when you enter the value e g FILTER PROCESSNAMES RTserver Specify FILTER ACCOUNTNAMES to export all process records with the accountname matching any of the specified strings Specify FILTER WORKLOADNAMES to export all process records associated with any of the specified workloads This filter is valid only if the CLASSIFY_BY qualifier is used to specify a classification scheme for your workload data UICS uic PIDS pid VOLUMENAMES string Specify FILTER UICS to export all process records with the UIC matching any of the specified UICs An asterisk may
5. Less inclusive keyword values override more inclusive values For example MONDAY 10 12 overrides EVERYDAY 8 17 for Monday but the Performance Manager selects data from 8 00 a m to 5 00 p m for all of the other days of the week For example ADVISE PERFORMANCE GRAPH _ SCHEDULE NOEVERYDAY WEEKDAYS 8 12 13 17 Graphs do not depict the time periods deselected by the SCHEDULE qualifier SELECT GREATER_THAN percent LESS THAN percent NOSELECT Use SELECT in conjunction with the optional threshold values which may be specified on a per graph type basis 198 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH If this qualifier is present before a graph is produced a check is made to see if the values to be graphed fall within the threshold values for the indicated percentage of points If so then the graph or pie chart is produced If not no graph is produced For details on THRESHOLD see the TYPE qualifier Keyword Meaning GREATER THAN percent At least percent of the graph points plotted must be greater than or equal to the threshold value specified with the TYPE qualifier LESS THAN percent At least percent of the graph points plotted must be less than or equal to the threshold value specified with the TYPE qualifier These keywords accept a single value representing the percentage of the points plotted that must meet the threshold criteria before the graph
6. Paged memory The total paged memory is expressed in pages and as a percentage of total memory and is represented by the PFN database This memory is consumed by the user working sets the system working set and the page cache free and modified page lists System working set The number of pages set by the system parameter SYSMWCNT is expressed in pages and as a percentage of paged memory User memory The difference between the amount of paged memory and the system working set paged memory system working set These are expressed in pages and as percentages of both paged and total memory It represents the memory available for user working sets and the free and modified page lists Chapter 3 Evaluate Performance in Detail 73 Tabular Report Sections Summary Statistics Sections The following example shows the Summary Statistics sections Tabular Report YQUEM VAX 6000 440 PA Vx x Tuesday 26 JAN 2006 09 00 to 10 00 FR AR EK A A 2 2 ok EE E E E E E K OK k k kk kk Node YQUEM Final Statistics 2 I 2 2 2 2 2K 2 2 2 2 2 k 2 2K k k k kkk aaahh Data Analyzed from 26 JAN 2006 09 00 00 00 to 26 JAN 2006 10 00 00 00 BEAL AA AA AAAS AAA AAA AR ISAK Avg Process Memory Counts Memory Utilization Avg Mem CPU Swapper Counts Proc Balset Free Modify Total Paged User Modify Header Header Swapper Count Count Pages Pages MEMutl MEMutl MEMutl MEMutl InSWP OutSWP InSWP Out
7. Record Header see page 540 Version Data Record see page 541 Memory Statistics Data Record see page 541 CPU Statistics Data Record see page 543 Secondary CPU Statistics Data Record see page 544 Page Statistics Data Record see page 545 I O Statistics Data Record see page 547 XQP Statistics Data Record see page 548 System Communication Services Data Record see page 549 Lock Statistics Data Record see page 551 Device Statistics Data Record see page 553 Disk Statistics Data Record see page 553 Server Statistics Data Record see page 555 Process Metric Statistics Data Record see page 556 Record Header This record header appears at the beginning of each data record 14 JAN 1997 10 00 00 00 14 JAN 1997 10 02 00 00 SNOLPD PROC 120 0 Where PROC could be CPU DEVI DISK IO LOCK MEMO PAGE PROC SYST or XQP Example Data item Description Position Lengt h 14 JAN 1997 start time interval start time 2 23 10 00 00 00 14 JAN 1997 end time interval end time 29 23 10 02 00 00 SNOLPD system na cluster node name 56 15 me PROC data type type of data in this 75 4 record 120 0 seconds seconds of data in this 82 7 interval 540 Performance Manager Administrator Guide Version Data Record Version Data Record The version record appears as the first record in the data file Vx x REPORT LOG FILE Vx x CPD
8. 2 2 2222 sess 337 8 Performance Manager Administrator Guide Set the Thresholds and Ranges iuiuussiusslluslllsslls ssl na 338 Change the Colors and Patterns 2 sls nn 339 Chapter 10 Customize the DECwindows Motif Real time Display 341 Access the Panel Manager isses eer RR PREG UY HESPPEPRPYUG P GF PPP E Gd eee ei 341 Specify Actions on Panels 222 22 esee 342 Terminate the SESSION eos pana AN LEVE UIN AN ENE LS Ibo eL xU endo aba a 345 How You Edit the Panel Instruments 2 sls 345 Enable the Build Mode clce le LR ee EA EU EE C E E Eg 346 Modify the Instruments ss rs 346 How You Set the Panel Options 0000s cette s 366 Set the Panel Status 2 llla hr 367 Specify the Panel Background 2 22 sls 367 Specify a Panel Title 2 2 esas 368 Specify the Panel Node and Metric Instance Data 2 sees 369 Remove Panel Menu 2 en n nee AG NG NANG NA eh hh e 370 Save the Panel 2 4 23 4x23 aa ua osti ETI i DG AGAR sees GGG P AG DEVET PEP REE PEDIRA EELE 371 Close the Panel enrera a eroe ese us ae car ha use euentu ee eere Gana hat eerta peat ters 371 Chapter 11 Use the Character Cell Real time Display 373 Character Cell Display Functions sna 373 Prerequisites oe verevoreo oO eo NAKA EE T NA KE Se Maen ibe ake FA vb eee aad ee Mae TATA NAAT 373 Start the Character Cell Displays 2 ea 374 Control the Displays
9. 216 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE COMPUTE QUEUE Number of Processes in COM and COMO Plots the number of computable processes categorized by Network processes NETWORK COMQ Interactive processes INT COMQ Batch processes BATCH COMQ Detached processes DETACHED COMQ TYPE CPU MODES CPU Modes Plots the percentage of CPU time spent in the various processor modes Multiprocessor synchronization MP SYNCH User USER MODE Supervisor SUPER Executive EXEC Kernel KERNEL Interrupt Stack INTERRUPT TYPE CPU_UTILIZATION CPU Utilization Plots 6 metrics for percent CPU utilization Interrupt stack and MP Synch CPU_MP_INT Detached processes CPU_DETACHED Interactive processes CPU_INTERACTIVE Batch processes CPU_BATCH Network processes CPU NETWORK Other CPU OTHER This is the default graph type TYPE DECNET System Wide DECnet Traffic Plots the number of DECnet operations per second in terms of Arriving packets ARRLOCPk Departing packets DEPLOCPK Transit packets ARRTRAPK TYPE DISKS DISK I O Plots the disk operations per second categorized by User DISK USER Paging DISK SWAPPING Swapping DISK PAGING Chapter 6 Performance Manager Commands 217 ADVISE PERFORMANCE GRAPH TYPE FAULTS Page Fault Rate Plots the page fault rate per second and places the rate into these categories Demand zero page faults DZRO
10. 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA SECO 0 8 10 2 1 6 0 1 2 9 0 0 76 1 12 6 3 0 2 8 3 Example Data item 0 8 sec interrupt 10 2 sec kernel 544 Performance Manager Administrator Guide Description of time in interrupt mode of time in kernel mode Position 91 100 Length 7 Page Statistics Data Record Example Data item Description Position Length 1 6 sec exec of time in executive 109 7 mode 0 1 sec super of time in supervisor 118 7 mode 2 9 sec user of time in user mode 127 7 0 0 sec compat of time in compatibility 136 7 mode 76 1 sec idle of time in idle mode 145 7 12 6 sec system c 9o of time in system mode 154 7 pu I K E 3 0 sec task cpu of time in task mode 163 7 S U C fill_1 Obsolete fill_2 Obsolete 2 sec_cpu_id CPU id number 190 3 8 3 sec busy wait of time in a busy wait 195 7 spintime Page Statistics Data Record The following example commands show how to export and display page statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP PAGE CLASS NODEFAULT PAGE _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP PAGE ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA PAGE 118 4 0 0 3 5 28 9 0 0 33 5 56 8 18 0 6 3 0 0 0 0 0 1 3 0 97 0 0 0 Example Data item Description Position Length 118 4 tot
11. DEADLOCK FIND DEADLOCK SEARCH INCOMING LOCKING LOCAL LOCKING OUTGOING LOCKING Sampled pool metrics IRP CNT IRP MAX LOCK CNT LRP CNT LRP MAX NP FREE BLOCKS NP FREE BYTES NP FREE LEQ 32 NP MAX BLOCK NP MIN BLOCK NP POOL MAX PG FREE BLOCKS PG FREE BYTES PG FREE LEQ 32 PG MAX BLOCK PG MIN BLOCK PG POOL MAX RESOURCE CNT SRP CNT SRP MAX Other system metrics SM RESPONSE MED RESPONSE LG RESPONSE CPU VUP RATING RELATIVE CPU POWER Select Up to 6 System Metrics CR for list DZROFAULTS GVALID 3 Enter an optional title for the CUSTOM graph 40 characters max Title PSPA CUSTOM GRAPH PAGEFAULTING 4 The following statements are keyed to the Prompting for System Metrics Graph 1 1 was entered to select System Metrics 2 CR was entered to select the list of system metrics Chapter 4 Generate Historical Graphs 129 Generate Custom Graphs 3 DZRO GVALID was entered to select these two metrics 4 Pagefaulting was entered to select the custom graph title PAGEFAULTING Rmo SINIY A QON LS Graph Process Metrics by User When graphing process metrics by user you must specify either of the following user names and process metrics m A list of up to six user names and a single process metric or m A single user name with a list of up to six process metrics Performance Manager graphs include the metrics for all users that match any one of the user names you specify in the list The resulting graph depicts
12. Example Data item Description Positi Length on Vx x image image version 92 7 REPORT LOG FILE report progr type of report 103 17 E am LOG FILE HISTO RY Vx x data vers version of data 122 7 file CPD input source file HISTORY or 133 22 1 LOGFILE Memory Statistics Data Record The following example commands show how to export and display memory statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP MEM CLASS NODEFAULT MEMORY _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP MEM ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA MEMO 160 0 327680 78 5 282821 44859 5000 277821 16352 0 70438 6 5 2 0 122 0 89 2 219 7 0 0 217 7 0 0 0 0 0 0 0 0 Example Data item Description Position Length 160 0 mbytes megabytes of phys 91 7 memory available 327680 pfn phypgcn pages of physical 100 7 t memory available 78 5 mem util percentage of total 109 7 memory in use 282821 paged of pgs for pageable 118 7 memory Appendix E Output Format for ASCII CSV Data 541 Memory Statistics Data Record Example Data item Description Position Length 44859 nonpaged of pgs for non paged 122 7 memory 5000 sys workset of pgs for the system 131 7 workingset 277821 user workset of pgs avail for user 140 7 workingsets 16352 0 modifyp size of the modified 149 7 page list
13. N w Oooh OODOoo C QJ tO C 00 O0 O0 H tO CO YQUEM DFSC7101 P i a A i E E Unknown YQUEM DFSC7102 A i i a P i Unknown YQUEM DFSC7103 3 B I Unknown YQUEM DFSC7104 5 s i Unknown YQUEM DFSC7105 i A s s i i i Unknown User Command ADVISE PERF REPORT TAB SECT DISK NODE YQUEM BEG 26 JAN 1997 09 00 00 00 END 26 JAN 1997 10 00 00 00 0UT TAB DISK YQUEM The following statistics are given for each mounted disk Work Avail 9 o The percentage of time that the disk had any type of I O request outstanding Paging The percentage of the Work Avail time that the disk was doing paging I O including I O to and from paging files image activations or global section writes Swping The percentage of the Work Avail time that the disk was doing swapping I O This includes both swapping and modified page writing Contlr 9 o The percentage of this disk controller s activity that was busy because of this disk 88 Performance Manager Administrator Guide Tabular Report Sections Rate s The number of I O operations per second performed by the disk Read Yo The percentage of read I O performed to the disk Remote I O The percentage of I O to a disk performed on behalf of other nodes in a cluster system Resp Time ms The response time is the mean time per I O request for the device in milliseconds This is the total time taken to perform an I O request and is the sum of the q
14. Other files that incurred the highest response times The metric graphed is FILE RESPONSE TIME TYPE TOP RESPONSE TIME IMAGES Top Image Response Time Plots the five plus Other Images images with the highest terminal response time The metric graphed is RESPONSE TIME TYPE TOP RESPONSE TIME USERS Top User Response Time Plots the five plus Other Users users with the highest terminal response time The metric graphed is RESPONSE TIME TYPE TOP RESPONSE TIME VOLUMES Top Response Time Disk Volume Plots the five plus Other Volumes disk volumes that have the highest response times The metric graphed is D RESPONSETIME TYPE TOP RESPONSE TIME WORKLOADS Top Workload Response Time Plots the five plus Other Workloads workloads with the highest terminal response time The metric graphed is RESPONSE TIME TYPE TOP RULE OCCURRENCES Top Rule Occurrences Plots the five plus Other rule identifiers that fired as a rate per hour The metric graphed is OCCURRENCES and is available only from history data TYPE TOP SPLITIO DISKS Top Split Operations Disk Device Plots the five plus Other disk devices that have the highest split I O operations The metric graphed is SPLITIO TYPE TOP SPLITIO FILES Top Split Operations Files Plots the five plus Other files that have the highest split I O operations The metric graphed is FILE SPLITIO TYPE TOP SPLITIO VOLUMES Top Split Opera
15. Pages Pages Alphabetic Alphabetic Numeric Alphabetic Alphabetic Numeric Numeric Next Panel 328 Performance Manager Administrator Guide CPU Queue Panel Descriptions Process Wait Panel Image CPU Panel Default Panel Descriptions The CPU Queue graph on the System Overview panel lets you access the following panels m Process Wait m Image CPU This series of panels lets you investigate the cause of high CPU queues on the monitored system You can investigate the top images using the CPU The following tables describe the information presented within each panel Label Metric Units Next Panel Computable Number of processes in Count Process Count COM queue with history Total Processes Number of processes Count scheduled Computable Number of processes on Count current COM queue Hibernating Number of processes in Count HIB state LEF Wait Number of processes in Count LEF state Page Fault Wait Number of processes in Count PFW state Top Images list CPU utilization of top Percentage Image CPU images Label Metric Units Next Panel CPU Utilization CPU utilization for this Percentage ES with history image CPU Utilization CPU utilization for this Percentage Es current Direct IO Rate image Direct I O for this image Operations per second Chapter 9 Use the DECwindows Motif Real time Display 329 Default Panel Descriptions Label Disk IO Rate Disk Th
16. Rate R This leftmost column is headed by the letter R and is an absolute value The value is not only given by the column height using the leftmost scale Rate scale but is also shown as a number at the bottom of the column The value is the number of direct I Os per second for the disk Chapter 11 Use the Character Cell Real time Display 389 Display RESOURCE Information Response R This rightmost column is headed by the letter R and is an absolute value The value is not only given by the column height using the rightmost scale Msec scale but is also shown as a number at the bottom of the column The value is the response time of the disk in milliseconds This is the average time to process one I O including both queuing time and service time Note that the bottom value in this column is the average number of I O packets in the disk queue If the value of the Rate or Response time exceeds 100 asterisks are shown at the top of the column but the value is still given at the bottom of the column If the actual value should exceed three digits then asterisks are shown in place of the value CPU Display The CPU display contains CPU statistics for analyzing a CPU resource limitation in a cluster wide manner An example of a resource CPU display is shown in the following table The top half of the CPU display contains a bar graph for each cluster node currently set for display The scale shown at right and left s
17. TYPE TOP PRCT FREE DISKS Top Percent Freeblks Disk Device Plots the top five plus Other disk devices in terms of percentage of free disk blocks The metric graphed is PRCT FREE TYPE TOP PRCT USED DISKS Top Percent Usedblks Disk Device Plots the top five plus Other disk devices in terms of percentage of used disk blocks The metric graphed is PRCT USED Chapter 6 Performance Manager Commands 225 ADVISE PERFORMANCE GRAPH TYPE TOP PRCT FREE VOLUMES Top Percent Freeblks Disk Volume Plots the top five plus Other disk volumes in terms of percentage of free disk blocks The metric graphed is PRCT FREE TYPE TOP PRCT USED VOLUMES Top Percent Usedblks Disk Volume Plots the top five plus Other disk volumes in terms of percentage of used disk blocks The metric graphed is PRCT USED TYPE TOP QUEUE DISKS Top Queue Disk Device Plots the five plus Other Disks disk devices that experienced the longest queue lengths The metric graphed is QUEUE TYPE TOP QUEUE VOLUMES Top Queue Disk Volume Plots the five plus Other Volumes disk volumes that experienced the longest queue lengths The metric graphed is QUEUE TYPE TOP READ DISKS Top Read Operations Disk Device Plots the five plus Other Disks disk devices that incurred the highest read I O rates The metric graphed is READIO TYPE TOP READ FILES Top Read Operations Files Plots the five plus Other files th
18. The Collection Definition dialog box lets you do the following actions m Modify create or remove collection definitions m Change collection parameters m Change collection schedules m Add or remove nodes Your changes to the Collection Definition file are applied when you pull down the Control menu and release on Save schedule file or exit If you change settings and then decide not to use the changes pull down the Control menu and release on Load schedule file to start over with the latest version When you are done with this window pull down the Control menu and release on Exit Before writing out a new schedule file or releasing the lock a message box appears with one of the following messages m If modifications have been made Do you want to write out your changes to the schedule file and release the lock m If no modifications have been made You have locked access to the schedule File do you want to release it to other users Press YES to release the lock Your modifications if any are written out Press NO to keep the lock Your modifications if any are not written out Modify or Create a Collection Definition The names of the collection definitions are listed in the box at the upper left hand corner To select a collection definition click the collection definition The definition and current parameters will appear Use the Tab key to move between parameter entries or point to the entry and click
19. m VUPS for SMP systems are computed by adding 85 percent threshold TD SMP VUP RATIO of the single processor VUPS rating for each additional processor If you decide that you would like to refine the scaling factors to more precisely reflect typical activity on your system then you need to know your system s hardware model ID number To get this enter the following at the DCL prompt n f getsyi hw model show symbol n Assume that this action returned the number 230 to you Then if you want to change the threshold for the number of jobs in the Compute queue from its default value of 1 30 to 1 45 you just add another line in MYRULES VPR in the following format Threshold TD COM SCALING 230 1 45 EndThreshold Change a Rule Literal Value Assume that security due to its additional overhead is not a priority for your systems Minimal security erasure I Os should occur on your system You would like to change the threshold used in rule I0180 from 1 0 to 0 1 You look in the source file for the factory rules PSPA EXAMPLES PSPA KB VPR and find TD HIGH ERASE IO 1 0 in the initial literals section Performance Manager Rule Literals This rule literals section is listed in the following table for ease of reference Literal TD HIGH ERASE IO 1 00 TD DISK QUEUE LENGTH 1 00 TD HIGH PROC PAGE FAULT 500 TD IO ERROR COUNT 2 TD 10 4 TD CPU 4 TD IMG 10 TD XQP CACHE 10 TD PAR 4 TD SYS 4 170 P
20. Datagram discard rate per second Message send rate Message receive rate NEN Block send rate ja o Block request rate HR Hn Kilobyte send rate HG N Kilobyte receive rate HG w Kilobyte map rate D Send credit waits 15 Buffered descriptor waits The following example shows the Performance Evaluation report SCS Statistics Performance NODE02 VAX 11 78 Evaluation Wednesday 14 JAN 2006 00 00 to 23 59 The following table gives the SCS message and transfer rates from this node to all other nodes in the cluster Values are rates sec CLUSTR PATH CIR DGS DGS DGS MGS MGS BLKS BLKS KB KB KB CRD BDT NODE PORT CUIT SND RCD DSD SND RCVD SEND RQSTD SEND RCVD MAPPED WAI WAI 1 2 3 1 DEFEND PAAO CI 0 NODE3 PAAO CI 0 AGRIC PAAO0 CI 0 SUPPLY PAAO CI 0 I 0 Total PAAO C Chapter 3 Evaluate Performance in Detail 53 Performance Evaluation Report Performance CLUSTER Evaluation Wednesday 14 JAN 2006 00 00 to 23 59 The following table gives the SCS message and transfer rates to the target node from other nodes in the cluster Values are rates sec TARGET NODE DEFEND SOURCE PATH CIR DGS DGS MGS MGS BLKS BLKS KB KB KB CRD BDT PORT CUIT SND RCV DSD SND RCVD SEND 0 0 0 0 0 1 0 0 0 0 0 1 Interpret cluster wide Lock Statistics The lock statistics are the first cluster wide statistics and follow the system specific SCS statistics To display only the lo
21. For example if you wanted to generate graphs and reports for the all Mondays in January but wanted to substitute a Tuesday for the Martin Luther King holiday you could create a text file with the following entries 04 JAN 2006 10 00 04 JAN 2006 12 00 04 JAN 2006 14 00 04 JAN 2006 16 00 11 JAN 2006 10 00 11 JAN 2006 12 00 11 JAN 2006 14 00 11 JAN 2006 16 00 19 JAN 2006 10 00 19 JAN 2006 12 00 19 JAN 2006 14 00 19 JAN 2006 16 00 25 JAN 2006 10 00 25 JAN 2006 12 00 25 JAN 2006 14 00 25 JAN 2006 16 00 The above date list indicates that the hours of 10 to 12 a m and 2 to 4 p m should be processed for the 4 days specified Tuesday the 19th is processed in place of Monday the 18th a holiday Release on the Modify button and a Performance Manager Dates File Selection box appears Performance Dates File Selection Load date ranges from file DATES TXT I OK Cancel Chapter 8 Use the DECwindows Motif Interface 271 How You Select Data for Analysis Enter the name of the file containing the date ranges and release on OK When the dates are successfully loaded the box is removed and the Start Time and End Time buttons are updated and desensitized Click the Calendar button to turn off the Dates File Filtering Options Click Filter to select a subset of data for reporting and graphing Click Modify to change the entries for filtering The Selection Filters dialog box lets you select a subset of the daily or hist
22. Source Start Time 26 JAN 1997 09 00 End Time 2 amp JAN 1997 10 00 Explanation The following table gives tr CS message and transfer rates from a source node to all other nodes in the cluster Values are rates IS MIS MIS BLKS Bits KB 050 SNO RCVO SEND ROSTO SEND To navigate through the Performance Evaluation Report click the Statistics label An option menu appears listing all of the Performance Evaluation Reports Release on the report section you want displayed You can also advance through the report by clicking on the appropriate direction arrows The Section label here labeled SOURCE changes according to the report being viewed and will be desensitized if irrelevant The arrows on either side of this label allow you to advance to other sections of the report 280 Performance Manager Administrator Guide How You Display Analyzed Data Press on the Node button and an option menu appears listing the nodes for which the report is available Release on the node you want displayed Only pool statistics are available by node To exit from the Performance Evaluation Window pull down the File menu and release on Close To customize the Performance Evaluation Report pull down the Customize menu and release on the menu item you want The Customize menu lets you alter the hotfiles report Set hotfile limit Release on the Set Hotfile limit menu item to display the Set Hotfile limit dialog box contai
23. Specifies a command string of less than 132 characters that is to be executed in the context of the created subprocess When the command completes execution the subprocess terminates and control returns to the parent process Chapter 7 Use Command Mode Commands 255 EXIT Description The SPAWN command creates a subprocess of your current process with the following attributes copied from the parent process All symbols except RESTART SEVERITY and STATUS Key definitions The current keypad state The current prompt string All process logical names and logical name tables except those explicitly marked CONFINE or those created in executive or kernel mode Default disk and directory Current SET MESSAGE settings Current process privileges Control and verification states Note that some attributes such as the process s current command tables are not copied EXIT The EXIT command returns you to the DCL command level Format EXIT Execute Procedure Format file spec Description The command causes subsequent commands to be obtained from the specified file instead of the user s terminal When all the commands are executed command input is returned to the user s terminal 256 Performance Manager Administrator Guide Chapter 8 Use the DECwindows Motif Interface This chapter provides information about using the Performance Manager DECwindows Motif Interface to perform Performance Manager analys
24. TYPE TOP USER PROCESSOR Top User Mode Processor Plots the five plus Other processors in terms of time spent in User mode as a percent of CPU time The metric graphed is P USER TYPE TOP USER VOLUME IO Top I O Users and the Disk Volumes they access Plots the top five plus Other user and volume name pairs in terms of their I O rate The metric graphed is USER VOLUME IO TYPE ZTOP WORKLOAD IMAGE ACTIVATIONS Top Image Activations Workload Plots the top five plus Other workloads in terms of image activations per second The metric graphed is IMAGE ACTIVATIONS TYPE TOP WRITE DISKS Top Write Operations Disk Device Plots the five plus Other Disks disk devices that incurred the highest write I O rates The metric graphed is WRITIO TYPE TOP WRITE FILES Top Write Operations Files Plots the five plus Other files that incurred the highest write I O rates The metric graphed is FILE WRITIO TYPE TOP WRITE VOLUMES Top Write Operations Disk Volume Plots the five plus Other Volumes disk volumes that incurred the highest write I O rates The metric graphed is WRITIO Chapter 6 Performance Manager Commands 229 ADVISE PERFORMANCE GRAPH Examples TYPE TOP VA IMAGES Top VA Space Images Plots the top five plus Other Images images that had the largest combined virtual address space by image name The metric graphed is VASIZE TYPE TOP VA USERS Top VA Space Users Plots
25. Tabular Report Sections Tabular Report PA Vx x Tuesday 26 JAN 1997 09 00 to 10 00 FRR EK A A EE E E K E k E k OK k BK OK kk koo k kkk k 29400201 29400206 29400209 2940020A 29401EFB 29401EFC 29401EFE Process SWAPPER Swapper CONFIGURE DSA111 SYS2 ERRFMT DSA111 SYS2 CACHE SERVER DSA111 SYS2 DSA111 SYS2 OPERATOR DSA111 SYS2 BHAT 2 DSA111 SYS2 DOS 34061 DSA111 SYS2 YQUEM VAX 6000 440 Node YQUEM Final Statistics FERRE EIA Data Analyzed from 26 JAN 1997 09 00 00 00 to 26 JAN 1997 10 00 00 00 FEAEEAALEDESARERE MASE SYSCOMMON 1 6 SYSCOMMON 1 4 SYSCOMMON SYSCOMMON 1 6 SYSCOMMON 310 57 SYSCOMMON 300 311 SYSCOMMON Image CPUtime DirI O BufI O Pg Flt FltI O Pri State Count min sec sec sec 16 16 HIB 10 8 HIB SYSEXE CONFIGURE 7 7 HIB SYSEXE ERRFMT 16 16 HIB SYSEXE FILESERV SYSEXE LOGINOUT 4 4 CUR SYSEXE LOGINOUT 4 4 CUR SYSEXE COPY 4 4 CUR SYSEXE LOGINOUT User Command ADVISE PERF REPORT TAB SECT PROC NODE YQUEM BEG 26 JAN 1997 09 00 00 00 END 26 JAN 1997 10 00 00 00 0UT TAB PROC YQUEM Extended Process Metrics The following example shows process metrics data Extended process metrics include these statistics PID The process identification number in hexadecimal User Name The user name for the process Account The account name for the process Globi MIN AVE MAX The
26. The following statements are keyed to the columns in the previous example 1 Explanatory node status 2 List of node names in the cluster 3 Number of hours spanned by the reporting period This histogram reflects a reporting period of 24 hours Chapter 3 Evaluate Performance in Detail 69 Histograms Combined CPU Usage Chart Histogram The following example shows a combined CPU usage chart histogram Histograms CLUSTER Page 17 PA Vx x Friday 18 SEP 2005 00 00 to 23 59 Legend 1 NODEO1 Cluster Utilization SUPPLY b DEMAND Scaled by CPU speed COMBINED CPU USAGE CHART utilized 2 Mg Dna DA racio 0000000 sc auiisiacioienss 00 aaaaa000a 0 4 00 bbbbb000000000000000bbbbbbbbbbbbb bbbbb bbbbbb 000000bbb0 00 000 bbbbbbbbcccccbb bbb bb bbbb 0 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 5 Each Column represents approximately 20 minutes starting from 18 SEP 00 00 00 to 18 SEP 23 59 00 An N indicates NO DATA The following statements are keyed to the previous example 1 Explanatory list of symbols in the histogram columns 2 Percentage utilizations of the most and least used nodes in a cluster for a requested period Imbalance in cluster wide CPU use represented by periods 4 cluster wide averages designated by O for all of the nodes included in the report Os print on top of all other symbols when coincident with another legal symb
27. This contains a value representing the sum of values of the average operations rate for all the current communications subrecords which were selected by the most recent COMMUNICATION_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains COM COMO_PROCESSES_ARE_AT_BPRI Derived This contains a Boolean value zero or one one true representing the presence of processes in computable outswapped state and running at base priority on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 422 Performance Manager Administrator Guide Performance Manager Data Cells COMPATMET F COMPAT Average percentage of CPU time spent in Compatibility mode for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP COMPUTABLE PROCESSES Derived This contains a value representing the number of computable processes MET F COM MET F COMO on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP COMPUTABLE PROCESSES OVR DEFPRI Derived This contains a value representing the number of computable processes with a scheduling priority at or above the default priority DEFPRI on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP COM_SCALING Derived This contains a valu
28. Utilization A usergroup workload family is a set of user based workload definitions The percentage of a resource s capacity being used 574 Performance Manager Administrator Guide VAXcluster system VUP Wait time Workload Workload definition Workload family A multipurpose system configured by interconnecting or clustering VAX processors and storage controllers to provide increased capabilities for sharing data distributing workloads and providing greater system and data availability VAX unit of processing speed The VUP rating measures the CPU power of a system compared to a VAX 11 780 A VAX 11 780 has a VUP rating of 1 0 Wait time or queueing time is the average time each request spends waiting in a queue for service During this time the request accomplishes no useful work Wait time may be derived specifically for each device in the system or for the system as a whole A representation of the actual system s resource demands Performance Manager software reports graphs and capacity plans based on workloads Workload definitions specify to Performance Manager software how to organize the system load into workloads A workload definition can be one or more users image names or processes that represent units of work on the system Workload definitions are identified and stored in the Performance Manager parameters file The goal is to express the system s total workload in manageable and meaningful units agai
29. cece ANA KANA hes 413 Appendix D Estimate Virtual Memory Needs 535 How Performance Manager Uses Virtual Memory 2 2 2222 535 FOGGING ose ube eee pes ce hae we apn AA wn ea een ea ea ea pee maa nama 536 10 Performance Manager Administrator Guide gear rm 537 For Integrity Servers and Alpha Systems 2222 sls 537 Appendix E Output Format for ASCII CSV Data 539 Record Header ossresppeneRerRee rp n REG EGG eg Pn deg ng e ng E pede gn gen ing degere deg ng 540 Version Data Record 222 22 541 Memory Statistics Data Record ess 541 CPU Statistics Data Record 2 lese 543 Secondary CPU Statistics Data Record 2 22 22 ssa 544 Page Statistics Data Record sess ls 545 I O Statistics Data Record 2 547 XQP Statistics Data Record 2 lesse 548 System Communication Services Data Record 2 sss 549 Lock Statistics Data Record essere 551 Device Statistics Data Record 2 elles 553 Disk Statistics Data Record 2 2 sss 553 Server Statistics Data Record 2 555 Process Metric Statistics Data Record esses 556 Appendix F How You Graph Seven or More CPUs 559 Step 1 Create a CSV file esas 559 Step 2 Create More CSV Files as Necessary 2 222 e 560 Step 3 Create a Single CSV File 22 22 ea 560 Resulting File After Merge and Edits 222 22 561 Step 4 Send the
30. 0 0 408 0 408 0 408 0 I MACNEIL CC Y4Y 156 0 156 0 156 0 252 0 252 0 252 0 1024 0 4096 0 16000 0 4620 0 4620 0 4620 0 1 DUA3 SYS5 SYSCOMMON SYSEXE SET Example Data item Description Position Length 26E00824 PID Process ID 91 8 MACNEIL proc name process name 102 16 00750 000021 asc uic ASCII UIC 121 14 4 priority software 138 2 priority LEF proc state process state 144 5 0 0 image count image count 152 7 0 0 cpu time CPU time in 161 7 minutes 0 0 dir io direct i o 170 7 rate second 0 0 buf io buffered i o 179 7 rate second 0 0 faults page 188 7 faults second 0 0 fault io page fault 197 7 i os second 408 0 min ws minimum 206 7 working set size 556 Performance Manager Administrator Guide Example 408 0 408 0 I MACNEIL CC Y4Y 156 0 156 0 156 0 252 0 252 0 252 0 1024 0 4096 0 16000 0 4620 0 4620 0 4620 0 Data item ave ws max ws mode user name account min gbl pgs ave gbl pgs max gbl pgs min prv pgs ave prv pgs max prv pgs ws default ws quota ws extent min virt ave virt max virt Process Metric Statistics Data Record Description average working set size maximum working set size process mode interactive batc h network deta ched other user name account minimum global pages average global pages maximum global pages minimum private
31. 1 2 3 4 ul Edit the first file Open the second file Append the CPU ids to the end of the Time record Append the CPU data for each collection interval to the end of the record for that collection interval Repeat for any additional files Insert a second comma after Time in the time record Edit the format of the time fields for best display in the Excel graph Change the quoted string to comma separated values For example change 15 JAN 2008 13 00 To 15 JAN 2008 13 00 For example for two files with a total of 12 CPUs 560 Performance Manager Administrator Guide File 1 File 2 Step 3 Create a Single CSV File PSPA CUSTOM GRAPH Node RX8620 Date 15 JAN 2008 13 00 13 10 Metric Values are Stacked eg Added to the left Units PROCESSOR Total Busy Time RX8620 0 RX8620 1 RX8620 2 RX8620 3 RX8620 4 RX8620 5 15 JAN 2008 13 00 0 9183 0 2475 0 1250 0 1250 0 1108 0 0033 15 JAN 2008 13 02 0 9192 0 2450 0 1233 0 1233 0 1100 0 0025 15 JAN 2008 13 04 0 9192 0 2458 0 1250 0 1250 0 1058 0 0008 15 JAN 2008 13 06 1 1150 0 3717 0 2358 0 2208 0 1892 0 0175 15 JAN 2008 13 08 0 9175 0 2433 0 1242 0 1242 0 1067 0 0000 PSPA CUSTOM GRAPH Node RX8620 Date 15 JAN 2008 13 00 13 10 Metric Values are Stacked eg Added to the left Units PROCESSOR Total Busy Time RX8620 6 RX8620 7 RX8620 8 RX8620 9 RX8620 10 RX8620 1
32. 9 0 0 9 0 2 BOOOCOCOC o rFPoOoooocoooo VMOUNT 1 VMSHELP 4 VTX CLIENT C 3 Totals 197 11 Chapter 3 Evaluate Performance in Detail 35 Performance Evaluation Report Performance YQUEM VAX 6000 440 Page 5 Evaluation PA Vx x Tuesday 26 JAN 1997 09 00 to 10 00 The following table summarizes the workload characteristics on a per image activation basis Note that values would be zeros if total number of image activations is zero PrimaryKey Avg Avg Avg Avg Mode of Soft Hard Direct Buff d Avg Images Secondary Key processes WSiz flts flts I0 I0 Cputim per activ inact image image image image image image Second INTERACTIVE 830 8 15 7 122 8 BATCH 0 0 0 0 0 0 NETWORK 367 4 4 6 35 0 DETACHED 1855 1592 0 9 0 5295 3 15542 The following statements are keyed to the columns in the previous example 1 By default the Performance Manager displays the process information by image name In this example images running or waiting on the system during the report time period are shown 2 Number of times that an image was activated during the report time period If an image has zero activations than it has been activated previously before the reporting period If you specify the secondary key as USERNAMES the Performance Manager displays the number of image activations per user 3 Number of soft page faults incurred by an image during the report time period divided by the number of activations If
33. CW DISK CHANNEL RATIO Derived This contains a percentage value representing the cluster wide ratio of disk operations for the current disk to the operations on the HSC s K SDI channel associated with the current disk for all nodes during the current interval Data Type NUMERIC Domains CLU CW DISK ERROR COUNT Derived This contains a value representing the cluster wide disk error count for the current disk for all nodes during the current interval Data Type NUMERIC Domains CLU CW DISK IO RATE Derived This contains a value representing the cluster wide disk operations rate for the current disk for all nodes during the current interval Data Type NUMERIC Domains CLU CW DISK THRUPUT RATE Derived This contains a value representing the cluster wide disk throughput rate in bytes per second for the current disk for all nodes during the current interval Data Type NUMERIC Domains CLU CW TOP FILE NAME Derived This contains a text string representing the file name of the file with the highest disk operations rate for all nodes for the current disk Data Type STRING Appendix C Performance Manager Data Cells 429 Performance Manager Data Cells Domains CLU CW TOP FILE OPCNT Derived This contains a value representing the disk operations rate of the file with the highest disk operations rate for the current disk for all nodes during the current interval Data Type NUMERIC Domains CLU
34. CW VOLUME NAME Derived This contains a string of text representing the cluster wide volume name for the current disk during the current interval Data Type STRING Domains CLU DATAGRAMS DISCARDEDSCS F DGDISCARD This contains the value representing the number of datagrams discarded per second by the local node and received from the remote node for the current configuration record and interval Data Type NUMERIC Domains CFG DATAGRAMS DISCARDED TALLY Derived This contains the sum of the values representing the number of datagrams discarded per second by the local node and received from the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG DATAGRAMS_RECEIVEDSCS_F_DGRCVD This contains the value representing the number of datagrams received per second on the local node from the remote node for the current configuration record and interval Data Type NUMERIC Domains CFG 430 Performance Manager Administrator Guide Performance Manager Data Cells DATAGRAMS RECEIVED TALLY Derived This contains the sum of the values representing the number of datagrams received per second on the local node from the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY
35. Domains CFG BLOCK_SEND_DATAS_INIT_TALLY Derived This contains the sum of the values representing the number of block transfers initiated per second on the local node to the remote node for all the current configuration subrecords that were selected by the most recent CONFIGURATION_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG BUFFERED_IO_RATEMET_F_BUFIO Average buffered I O rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BUFFER_DESC_QUEUE_RATESCS_F_QBDT_CNT This contains the value representing the number of times per second that a block transfer was queued because there were no available buffers on the local node to receive data from the remote node for the current configuration record and interval Data Type NUMERIC Appendix C Performance Manager Data Cells 417 Performance Manager Data Cells Domains CFG BUFFER DESC QUEUE TALLY Derived This contains the sum of the values representing the number of times per second that a block transfer was queued because there were no available buffers on the local node to receive data from the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG BYTES_FREE_IN_NONPAGED_POOLMET
36. Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PFCDEFAULTPAR F PFCDEFAULT The value of the SYSGEN parameter PFCDEFAULT which sets the page fault cluster size for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 512 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN PFRATHPAR F PFRATH The value of the SYSGEN parameter PFRATH which specifies the page fault rate above which the limit of a working set is automatically increased for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PFRATLPAR F PFRATL The value of the SYSGEN parameter PFRATL which specifies the page fault rate below which the limit of a working set is automatically decreased for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PHYSICALPAGEPSPAR F PHYSICALPAGES The value of the SYSGEN parameter PHYSICALPAGES which sets the maximum number of physical pages to be used for the local node for the current interval record in LOCAL domain and for the local node for the l
37. Domains COM CFG CPU DSK FIL LOC PRO TAP FILE_DEVICEFIL_A_DEVICE This contains a string indicating the disk device on which the current file in the hot file subrecord is located Data Type STRING Domains FIL FILE_DIRECTORYFIL_A_DIRECTORY This contains a string indicating the disk directory in which the current file in the hot file subrecord is located Data Type STRING Domains FIL FILE_EXTENT_CACHE_AR Derived Attempt rate per second to the extent cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP FILE_EXTENT_CACHE_HR Derived Hit ratio to the extent cache for the local node for the current interval record Calculated by dividing the number of extent cache hits by the number of extent cache attempts hits misses times 100 Data Type NUMERIC Appendix C Performance Manager Data Cells 447 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP FILE HEADER CACHE AR Derived Attempt rate per second to the file header cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP FILE_HEADER_CACHE_HR Derived Hit ratio to the file header cache for the local node for the current interval record Calculated by dividing the number of file header cache hits by the number of file header cache attempts hits misses times 100 Data Type NUMERIC Domain
38. Each node name begins a line of summary statistics as in the case of By Cluster format These statistics show the contribution of each individual node to each summary statistic Chapter 3 Evaluate Performance in Detail 95 Tabular Report Sections By Cluster or By Node Summary Statistics The following is an explanation of memory statistics in By Cluster and By Node formats Total MEMutl This is equal to Total Memory Free Pages Total Memory Memory Queue avg The average number of processes waiting for available memory The value of Memory Queue is equivalent to the count of processes in the computable outswapped queue COMO Proc Count avg The number of processes in the system including SWAPPER Balset Count avg The number of processes resident in the balance set Page Faults sec The total number of page faults per second during the example interval Hard Faults 9 7 The percentage of Page Faults that required a read from disk This is Read I Os Page Faults expressed as a percentage Soft Faults 9 The percentage of Page Faults that were resolved from memory that is without reading from disk This is equal to 100 Hard Faults 9o Gvalid Faults sec The number of page faults per second resolved as valid global pages already in memory during the example interval System Faults sec The number of page faults incurred in system space per second during the example interval InSWP Coun
39. IMAGENAME REPORT Provide An 80 column report showing process CPU memory and IO counts on a per image activation basis The default primary key is MODE no secondary key A 132 column report showing process CPU memory and IO statistics primarily as rates Some UAF parameters are also provided The default primary and secondary keys are USERNAME and IMAGENAME A 132 column report showing primarily process memory related statistics Some UAF parameters are also provided The default primary and secondary keys are IMAGENAME and USERNAME A 132 column report showing primarily process IO related statistics as rates The default primary and secondary keys are USERNAME and IMAGENAME A 132 column report showing primarily process CPU related statistics The default primary and secondary keys are USERNAME and IMAGENAME Provide Group process statistics by the process mode Interactive Batch Network or Detached Group process statistics by the process s User name Group process statistics by the process s Image Name Group process statistics by the process s UIC Group Group process statistics by the process name Group process statistics by the workload name You must specify CLASSIFY BY to indicate the workload family that you intend to use Group process statistics by the process s account name Group process statistics by the process s PID Provide Process statistics detail lines by the process mode Interactiv
40. NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP INCOMING DEADLOCK MESSAGE RATEMET F DLCKMSGS IN This contains the number of incoming deadlock detection messages per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP INCOMING DIRECTORY FUNCT RATEMET F DIR IN This contains the number of incoming directory operations per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP INCOMING_LOCK_CONVERSION_RATEMET_F_ENQCVT_IN This contains the number of incoming enqueue lock conversion requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP INCOMING_LOCK_DEQUEUE_RATEMET_F_DEQ_IN This contains the number of incoming dequeue lock requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP INCOMING_LOCK_ENQUEUE_RATEMET_F_ENQNEW_IN This contains the number of new incoming enqueue requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 458 Performance Manager Administrator Guide Performance Manager Data Cells INTERACTIVE COUNTMET F INTERACTIVE This contains a value representing the average number of interactive processes on the local node for the current interval Data Type NUMERIC Domains COM CFG C
41. Page Faults Per Second If metrics of differing units coexist on the graph the y axis label is blank Y Axis Markers Axis markers indicate the magnitude and time of any point on the graph The x markers indicate the time and are displayed differently depending on the graphic format The x markers are displayed as HH MM HH MM SS MM DD or MMM depending on the graph time range and graphic format When graphing historical data month or even years may appear depending on the time range selected The y markers are based on the maximum value of all the data points The increments are obtained from an internal table to make the graph easy to read You may define the maximum value on the y axis by using the Y AXIS MAXIMUM keyword Legend The legend appears at the bottom of the graph for all graphs other than ANSI graphs The legend identifies the name of the metric and the color pattern or graph character associated with it For the predefined TOPxxx graphs the items are always in the same format and order for example other users workloads images volumes disks and so forth topmost second top third top fourth and fifth 112 Performance Manager Administrator Guide Composite Graphs Composite Graphs If multiple nodes contribute data to a graph the graph is considered a composite graph Data for each node can be added together scaled by some factor and added or averaged The method used depends on the metric bei
42. Percentage of time in supervisor mode for this physical processor Percentage of time in user mode for this physical processor Percentage of time in interrupt stack for this physical processor Percentage of time in compatibility mode for this physical processor Percentage of time in null mode for this physical processor SO 2004 SL SOY GUT FB puo INIG JE Percentage of time physical processor n multiprocessor synchronization mode for this HG o Number of Performance Manager data records contributing to the above statistics Interpret SCS Statistics If included in the Performance Evaluation Report SCS statistics follow the CPU mode statistics for each node in a cluster This report also presents cluster wide SCS statistics The SCS statistics are not included in the default group of Performance Report options To enable the SCS statistics display from the Performance Evaluation Report use the following qualifier INCLUDE SCS STATISTICS 52 Performance Manager Administrator Guide Performance Evaluation Report The following statements are keyed to the columns in the SCS Statistics report 1 Other nodes in cluster The Performance Manager reports SCS message and transfer rate statistics as seen by node NODE2 to the other nodes listed here Port name if multiple paths are possible Physical circuit connecting all nodes on a common interconnect Datagram send rate per second Datagram receive rate per second
43. Periodicity e 317 Phase V support logical name definition e 406 Physical Memory Utilization graph e 191 Pie charts components of e 286 custom e 295 display of e 286 293 selecting e 288 top system use e 294 Playback data displays e 178 337 Predefined graph generate all e 191 245 identify the type e 111 types e 191 Primary keys e 34 Proc Count e 75 Process data options for reporting e 34 faults panel e 332 metric e 90 for graphs e 126 generate graphs by user in command mode e 130 generate graphs by user in prompt mode e 130 statistics window e 282 wait panel e 329 PROCESSES graph type e 191 PSPASDAILY COM e 16 PSPASDISPLAY PROCESS CPU UNNORMALIZE D logical e 405 PSPASDNS NAMES logical e 406 PSPASEXAMPLES logical name described e 406 PSPASGETDATA COM file e 406 PSPASGIVE DEVICE SERVICE logical e 406 PSPASGRAPH CHARS logical e 406 PSPASGRAPH FILE DEVICE logical e 407 PSPASGRAPH FILE DIRECTORY logical e 407 PSPASGRAPH LEGEND FONT POINT logical e 407 PSPASGRAPH PATH logical e 407 PSPASHLS logical e 407 PSPASPIE FONT POINT logical e 408 PSPASPS RGB 1 through PSPASPS RGB 6 logicals e 408 PSPASSKIP DISK FILTER logical e 409 PSPASSKIP PIE PERCENT logical e 409 PSPASSUPRESS TAPE STATS BY VOLUME logical e 409 PSPASUNNORMALIZE CUSTOM CPU logical e 409 Q Qualifier NO STACK e 191 230 AVERAGE e 191 241 BEGINNING qualifier for Playback data displays e 178 to export data e 182 to generate graphs
44. STACK qualifier The THRESHOLD keyword specifies a threshold value associated with the graph The m specifier is a positive decimal value A horizontal line is placed on the graph at the position on the Y axis associated with the value You can use THRESHOLD in conjunction with the SELECT qualifier to prevent the generation of the graph or pie chart The Y AXIS MAXIMUM specifies a fixed y axis to be used for the graph The default behavior is to setup the y axis so that the maximum data point appears near the top of the graph This graph modifier allows you to specify the y axis so that you can compare data from different graphs without having different scales on the y axis The n specifier is a positive decimal value 200 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH You can specify multiple graphs in a single command For example you can specify TYPE TOP IO DISKS TOP HARDFAULTING IMAGES Of course TYPE ALL GRAPHS generate all of the predefined graphs To suppress a graph type specify NO graph type CPU UTILIZATION is the default graph type The following list contains all of the available Performance Manager graphs NO ALL GRAPHS NO CPU MODES CUSTOM NO DISKS NO FILECACHE NO LOCKS NO PROCESSES NO TERMINALS NO TOP BLKS R NO TOP BUFIO IMAGES NO TOP BUFIO WORKLOADS NO TOP BUSY PROCESSOR NO TOP CHANNEL IO NO TOP CHANNEL THRUPUT NO TOP COMPAT PROCESSOR NO TOP CPU RULE O
45. Specify FILTER IMAGENAMES to select all process records with the imagename matching any of the specified strings Do not specify any trailing EXE nor the file version device or directory FILTER PROCESSNAMES amp string Specify FILTER PROCESSNAMES to select all process records with the processname matching any of the specified strings The match string is case sensitive so if the process names have any lowercase letters spaces or tabs use double quotes when you enter the value e g FILTER PROCESSNAMES RTserver FILTERZACCOUNTNAMES Specify FILTERZACCOUNTNAMES to string select all process records with the accountname matching any of the specified strings FILTER WORKLOADNAMES Specify FILTER WORKLOADNAMES to workloadname select all process records associated with any of the specified workloads This filter is valid only if the CLASSIFY_BY qualifier is used to specify a classification scheme for your workload data Chapter 7 Use Command Mode Commands 243 SELECT Keyword Description FILTER UICS uic Specify FILTER UICS to select all process records with the UIC matching any of the specified UICs An asterisk may be used to wildcard either the group or user field of the specified UICs FILTER PIDS pid Specify FILTER PIDS to select all process records with the PID matching any of the specified PIDs FILTER VOLUMENAMES string Specify FILTER VOLUMENAME
46. TORREY VOBA Node Name YQUEM IMAGENAME CMS of Page Faults Avg of of Uptime Cputim activ per Actvtn Ws Direct Buffered of image image User ations Soft Hard size 1 0 I O Cputim sec sec Totals Node Name YQUEM IMAGENAME of Page Faults Avg of of Uptime Cputim activ per Actvtn Direct Buffered of image image ations Soft Hard i I O Cputim ARROYO BHAT FORD QUANG VERRIER Totals Chapter 3 Evaluate Performance in Detail 39 Performance Evaluation Report Performance YQUEM VAX 6000 440 Evaluation Tuesday 26 JAN 1997 09 00 to 10 00 The following table summarizes the workload characteristics on a per image activation basis Note that values would be zeros if total number of image activations is zero PrimaryKey Avg Avg Avg Avg Image of Soft Hard Direct Buff d Avg Images Secondary Key processes WSiz flts flts I0 10 Cputim activ inact image image image image 8 0 0 1 0 0 2 0 DEBUGSHR DECW BOOKREAD DECWSCLOCK DECWSDWT FONT F3 OO HOOD OG W DD DOO kho eoooooono DDDDDQD QD QO Interpret Process Statistics by Workload The following example illustrates the process statistics section displayed by workload This report displays process statistics for each workload definition in the workload family supplied with the CLASSIFY BY qualifier In this example the Performance Manager uses the workload family MODEL TRANSACTIONS including the workload definit
47. Tabular Report and Dump Reports You must specify the time period the nodes and the processing options desired before these functions can be performed To analyze Performance Manager data 1 Pull down the Select menu and release on the menu item you want 2 The Select menu lets you specify the following data for analysis m Data collected today m Specific data m Data collected during the last hour of the current day Select Today s Data By default the Performance Manager selects daily data for all nodes and analyzes data for the Analysis Performance and Tabular Reports and the graphs and pie charts unless you have adjusted these options by using the Specific Data menu item The Performance Manager uses 00 00 00 of the current day as the beginning time and the current time as the ending time and chooses the CPD collection as the source To select the default release on Select Today from the Select menu You can override the defaults if you have a local resource file PSPA SELECT DAT that sets your preferred defaults See the section Data to be Analyzed see page 268 for a discussion of the Resource file 266 Performance Manager Administrator Guide How You Select Data for Analysis Select Specific Data Choose the Specific Data menu item from the Select menu to specify data to be analyzed Performance Manager displays the Performance Manager Data Selection dialog box Performance Data Selection Data to be Anal
48. You may want to use both the DATES qualifier and the SCHEDULE qualifier together For example if you need a graph depicting an average Monday through Thursday prime time for the first week of September October November and December of 2005 your schedule and dates file would be as follows MYDATES DAT 07 SEP 2005 11 SEP 2005 05 OCT 2005 09 OCT 2005 Q2 NOV 2005 06 NOV 2005 07 DEC 2005 11 DEC 2005 To use this dates file m Enter a command similar to the following ADVISE PERFORMANCE GRAPH SCHEDULE NOWEEKENDS WEEKDAYS 9 12 _ 14 17 NOFRIDAY DATES MYDATES DAT AVERAGE WEEKLY Use Binary Graph Data The DECwindows and the command mode interfaces let you save the graph data that you selected for viewing into a file This binary file can later be reloaded into either of the interfaces and the graphs presented eliminating the need for re analyzing the data 114 Performance Manager Administrator Guide Components of Pie Charts The following example shows the command mode interface ADVISE PERFORMANCE PSPA gt SELECT GRAPHS BY NODE BEG 26 JAN 2006 9 END 26 JAN 2006 17 PSPA gt SAVE JAN26 GRAPHS PSPA gt EXIT ADVISE PERFORMANCE PSPA gt LOAD JAN26 GRAPHS PSPA gt GRAPH TYPE CPU_UTILIZATION NODE NODE1 You can modify the PSPA DAILY COM file in PSPASEXAMPLES to save the graph data during a nightly batch job and the next day load the data if the analysis report indicated any problems Compo
49. a hot file record is selected if accessed by the specified process When you specify filtering by disk device a hot file record is selected if located on the specified device For hot file records matching both process and disk device specify filtering by both process and device Process data can be filtered by using any of the filter keywords USERNAMES IMAGENAMES PROCESSNAMES ACCOUNTNAMES UICS PIDS or WORKLOADNAMES If a process record s identification information matches any of the identification specifications that are specified then that record is selected Likewise disk data can be filtered by using either the filter keywords VOLUMENAMES or DEVICENAMES If a device record s identification information matches any of the volume names or device names that are specified then that record is selected 232 Performance Manager Administrator Guide ADVISE PERFORMANCE PIE CHART The following table lists the FILTER keyword options Keyword Description USERNAMES string Specify FILTER USERNAMES to chart all process records with the username matching any of the specified strings IMAGENAMES string Specify FILTER IMAGENAMES to chart all process records with the image name matching any of the specified strings Do not specify any trailing EXE nor the file version device or directory PROCESSNAMES string Specify FILTER PROCESSNAMES to chart all process records with the process name matching any o
50. custom graphs quickly and view any item categories to see what choices exist Use the TITLE keyword to override the Performance Manager supplied title The text string may be a maximum of 40 characters The STACK keyword for a particular graph type overrides the setting established by the STACK qualifier The THRESHOLD keyword specifies a threshold value associated with the graph THRESHOLD does not apply to pie chart graphs and is ignored The m specifier is a positive decimal value The Y AXIS MAXIMUM specifies a fixed y axis to be used for the graph The default behavior is to setup the y axis so that the maximum data point appears near the top of the graph This graph modifier allows you to specify the y axis so that you can compare data from different graphs without having different scales on the y axis The n specifier is a positive decimal value You can specify multiple graphs in a single command For example you can specify TYPEZ TOP IO DISKS TOP HARDFAULTING IMAGES Of course TYPE ALL GRAPHS generates all of the predefined graphs CPU UTILIZATION is the default graph type For a list of valid graph types see the chapter Performance Manager Commands PIE CHART Use the PIE CHART command to produce a pie chart instead of a graph The PIE CHART option has the same format as the GRAPH command however the data is presented as a pie chart instead of as a graph Format PIE CHART 248 Performance Manager Administ
51. each unit of work is called a transaction In the context of a history file interval is an ADVISE EDIT ADD HISTORY qualifier and a Performance Manager parameter file element In the context of a daily CPD data file the interval for writing records is fixed at two minutes In the context of real time displays the interval is user defined default interval 10 seconds minimum 1 second The journal file is when Performance Agent software creates one history journal file for each history file These are used by the update process in conjunction with the daily data files to recreate corrupt or deleted history files The filename has the form PSDC DATABASE PSDC nodename_dd mmm yyyy_dd mmm yyyy name_JOU Do not delete these files If they exist they are needed Glossary 569 Knowledge base Measured workload Model The knowledge base is a file consisting of Performance Manager performance rules used to analyze daily or historic Performance Manager data It may be augmented using an auxiliary knowledge base The measured workload is the workload that can be observed on the system Ideally the measured workload is identical to the actual workload but specified in different terms The measured workload is specified by the resource demands it places on the actual system These demands or loadings are given as the service times of the users at each device along with the number of expected transactions at each device A model is an a
52. nw of workloads nd 3t of disks nf of hotfiles nuid of activity entries for users and images of specific disks nn number of nodes 1 set to 1 if BY NODE graph option is off 536 Performance Manager Administrator Guide For Reports For Reports For the Performance Evaluation Process Statistics and the Tabular Report Process Metrics section To estimate the process virtual pages needed for the process reports m Use the following formula Pages required for reporting number of process instances 0 43 The number of process instances is the number of unique occurrences of a process PID imagename username processname accountname UIC group and execution mode Each of these item keys may be disabled by selecting reports or report options that do not require that level Performance Manager provides a set of default processing options that is designed to provide a medium amount of detail for the reporting and graphing functions but that would require significantly less memory than if all processing options were enabled For Integrity Servers and Alpha Systems When selecting data for both reporting and graphing from either Command Mode or DECwindows To estimate the number of pages needed pagelets on Integrity and Alpha systems m Use the following formula Pages required for reporting Pages required for graphing 20000 A safe approach is to provide as much process PFGLQUOTA as possible for the
53. on behalf of that lock DLock Find The number of times a deadlock was found The system selects a victim of the deadlock and does not grant the new lock or lock conversion request Total Locks The total number of locks taken out on all resources This number is an average of the examples taken at each recording interval over the reporting interval Total Resrcs The total number of resources that can be locked This number is an average of the examples taken at each recording interval over the reporting interval System Communication Service Rates These statistics are collected from OpenVMS performance statistics for each node in a cluster that was present during the reporting interval Each line of statistics gives the name of the node that is sending data to or receiving data from the local node There are three types of messages Datagrams m Block transfers m Sequenced messages Chapter 3 Evaluate Performance in Detail 85 Tabular Report Sections Datagrams are used primarily by DECnet the CI and by the HSC for error logging The delivery and order of messages is not guaranteed For block transfer mode if I Os are targeted to disks on an HSC the Kbytes mapped by the local node are recorded but not the transfer counts nor the Kbytes transferred This is because the HSC actually initiates the block mode transfer If I Os are targeted to disks hosted on an OpenVMS node MSCP server the hosting OpenVMS node
54. 00 00 to 23 59 ANALYSIS SUMMARY for node SUPPLY Number of Records Processed Number of Records satisfying rule conditions Number of Records not satisfying rule conditions 687 3 Number of Conclusions The following statements are keyed to the rows in the previous Analysis Summary report 1 Number of Performance Manager records analyzed for the specified reporting period 2 Number of Performance Manager records that satisfied rule conditions This number does not necessarily equal the number of rules that fired due to rule threshold values Although a record may satisfy a rule condition the number of occurrences required to fire the rule may not be sufficient during the reporting period 3 Number of Performance Manager records that did not satisfy any rule conditions 4 Number of conclusions generated for the node being analyzed Although records may have satisfied a specific rule the number of occurrences of records satisfying a specific rule may not have reached the occurrence threshold This means the number of conclusions may be zero while the number of records satisfying the rule conditions is greater than zero 28 Performance Manager Administrator Guide Analysis Reports Cluster Rule Analysis Report The following example shows a Cluster Rule analysis report Full Analysis CLUSTER Saturday 01 0CT 1996 00 00 to 23 00 CONCLUSION 1 L0050 Queues are forming on heavily used disks Longer delays will be e
55. 100 Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DIRECTORY_INDEX_CACHE_AR Derived Attempt rate per second to the directory index cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DIRECTORY_INDEX_CACHE_HR Derived Hit ratio to the directory index cache for the local node for the current interval record Calculated by dividing the number of directory index cache hits by the number of directory index cache attempts hits misses times 100 Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DIRECT_IO_RATEMET_F_DIRIO Average direct I O rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 433 Performance Manager Data Cells DISK BUSY PERCENTDEV F BUSY This contains a value representing the average percent of time the I O requests were outstanding to the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK BUSY PERCENT TALLY Derived This contains the sum of the values representing the average busy percentage for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_CACHE_NAMEDEV_A_CACHENAME This contains a string represe
56. 100 maximum CPU Only CPU is busy and no disk Percentage I O is busy CPU amp I O Busy CPU is busy and disk I O Percentage is also busy I O Only Disk I O is busy when Percentage CPU is not busy CPU amp I O Idle CPU and I O are not busy Percentage Top Users list CPU Utilization by User Percentage User CPU Label Metric Units Next Panel Int Interrupt mode time Percentage MP MultiProcessor Percentage synchronization mode time Ker Kernel mode time Percentage Exc Executive mode time Percentage E Sup Supervisor mode time Percentage s Usr User mode time Percentage Cmpt Compatibility mode time Percentage E 326 Performance Manager Administrator Guide User CPU Panel Process CPU Panel Default Panel Descriptions Label Metric Units Next Panel CPU Utilization CPU utilization Percentage with history CPU Utilization CPU utilization Percentage m current Direct IO Rate Direct I O for this user Operations per second Disk IO Rate Disk I O for this user Operations per ss second Disk Thruput Disk throughput Kilobytes per a second Hard Fault Rate Hard page fault rate Faults per second Soft Fault Rate Soft page fault rate Faults per second Buf I O Rate Buffered I O rate Operations per second User Processes CPU utilization by process Percentage Process list for this user CPU User Images CPU utilization by image Percentage Image CPU list for this user Label Metric Units Nex
57. 308 Performance Manager Administrator Guide How You Customize Create a Workload Definition Use the following options to create workload definitions Workload name Enter the name of the workload definition The workload definition s name is limited to 20 characters Workload is unique by lets you specify a category for workload summarization A workload will be defined for each unique element of the category you choose Click the Workload is unique by toggle button and then click MB1 on the box beneath this that displays Account name when first accessed The option menu will appear displaying the following items m Account name m Process name m Image name m UIC group m User name m PID Click MB1 on the category you want Include these processes Set the toggle buttons to specify the processing modes to be included by the workload definition m Interactive m Batch m Network m Detached Process base priority Enter the minimum and maximum values for the process Base Priority to be included in the definition Values can range from O to 31 and the minimum value must be less than or equal to the maximum Workload is defined by sets of items Selects the alternative to unique by criteria lets you enter lists of user criteria or images or both for defining a workload Matching requirement This is an option menu with two entries Press MB1 on the current setting to view the choices Chapter 8 Use the DE
58. AUGMENT DEVICE DIRECTORY MY SITE KB PSDC EDIT EXIT ADVISE PERFORMANCE REPORT ANALYSIS In either case the Performance Manager reads in the compiled user rules in addition to the default Performance Manager rules when generating an Analysis Report ADVISE PERFORMANCE GRAPH HISTORY MONTHLY USER _ TYPE TOP CPU RULE OCC The top firing rules concerning CPU usage during the last month are displayed Chapter 6 Performance Manager Commands 177 ADVISE PERFORMANCE DISPLAY ADVISE PERFORMANCE DISPLAY Format Description Qualifiers The ADVISE PERFORMANCE DISPLAY command invokes dynamic displays using a DECwindows Motif interface or using a character cell terminal ADVISE PERFORMANCE DISPLAY display keyword Use the ADVISE PERFORMANCE DISPLAY command to produce dynamic Performance Manager displays The following section describes the qualifiers you can use with the ADVISE PERFORMANCE DISPLAY command to control displays The display keywords are m CHARACTER CELL Provides dynamic displays on a character cell terminal m WINDOWS Provides dynamic displays using a DECwindows Motif interface Some of the following qualifiers are valid for both the CHARACTER CELL and WINDOWS interfaces while others are valid for only CHARACTER CELL Each qualifier has one of the following interface annotations m C CHARACTER CELL only C W Both CHARACTER CELL and WINDOWS BEGINNING time C Specifies the starting date and tim
59. Actual or real workload The actual or real workload is the work actually performed by the computer system Ideally the actual workload is identical to the business workload Adapter An adapter is a hardware interface between a device controller and a system backplane or bus Analysis report An analysis report is a report that identifies the effects of system parameter settings hardware configurations and workload mixes on the performance of any cluster node or the entire cluster system This report provides conclusions with supporting evidence and makes recommendations Analysis summary The analysis summary is a short summary appearing in the analysis report after each node analysis The summary contains the following information Number of Performance Manager records analyzed for the specific reporting period Number of Performance Manager records that satisfied any rules Number of Performance Manager records that did not satisfy any rules Number of conclusions generated for the node being analyzed Archive Archive is the act of compressing Performance Manager daily data records into history files which can be used in place of daily data to create reports models and graphs Auxiliary knowledge base Auxiliary knowledge base is A collection of user defined site specific rules and thresholds that have been compiled with the rules compiler name KB and used to augment the factory supplied rules Auxiliary rules Auxiliary rule
60. An example of adding a new rule to check for disk space shortages follows 166 Performance Manager Administrator Guide Implement Changes To creating a new rule m Use the following code Rule UI101 Any disk full eq True Occurrences td I0 Evidence Volume name Disk most full x Disk free pages Disk most full x Time Conclusion The following disks are almost full You should purge or delete any unnecessary files You might also consider moving some files to a disk with more free space Brief conclusion Disk space shortage Clean up disk or off load some files EndRule You could insert the type of rule shown in the previous example in many possible places within the I O decision tree used by the Performance Manager software which would add more conditions to those presented here Also you might have a large reference database that completely fills up N 1 volumes of its volume set so you would want to exclude those from consideration Finally you might want to show additional data cells in the Evidence Change a Threshold Value Assume that you would like to be even more proactive detecting potential bottlenecks on disks You might want to reduce the global threshold on disk queue length which is used as an initial condition in many disk rules from 1 0 to 0 66 and see how many more rule firings you get as a result of this change Chapter 5 Customize the Knowledge Base 167 Implement Changes The lis
61. BUDNN OU Oc C C NJ C CO Cc 2 C Oc OD DN O N BDWOOUOO Hr ON The following statements are keyed to the columns in the previous report 1 Device identification of path to remote SCS nodes 2 Node name of the cluster node 3 Adapter type for the cluster node all Ethernet adapter types are reported as NI Number of datagrams per second sent and received by this node s port Number of messages per second sent and received by this node s port Number of disk IOs per second delivered through this port Number of kilobytes per second delivered through this port to disks Go al ce Ur cm Number of kilobytes per second total delivered through this port Interpret cluster wide Disk Statistics The disk statistics follow the cluster wide CI NI and adapter statistics To display only the disk statistics section of the Performance Evaluation Report use the following qualifier INCLUDE DISK STATISTICS To disable the disk statistics display from the report use the following qualifier INCLUDE NODISK STATISTICS 56 Performance Manager Administrator Guide Performance Evaluation Report The following report shows an example of the disk statistics section of the Performance Evaluation report Performance CLUSTER Page 8 Evaluation PA Vx x Wednesday 14 JAN 2006 13 00 to 14 00 The following table gives the summary of all disk activity as seen by the indicated node An for service node indicates that more than
62. C Performance Manager Data Cells 505 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN DORMANTWAITPAR F DORMANTWAIT The value of the SYSGEN parameter DORMANTWAIT which indicates the number of seconds that may elapse without a significant event before the system treats a low priority computable process as a dormant process for scheduling purposes for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN FREEGOALPAR F FREEGOAL The value of the SYSGEN parameter FREEGOAL which establishes the number of pages that you want to reestablish on the free page list following a system memory shortage for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN FREELIMPAR F FREELIM The value of the SYSGEN parameter FREELIM which sets the minimum number of pages that must be on the free page list for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN GBLPAGEPSPAR F GBLPAGES The value of the SYSGEN parameter GBLPAGES which is the global page table entry count which establishes the size of the global page table and the limit for the to
63. CPU utilization threshold on the strip chart in the CPU panel is set to alert you when CPU utilization exceeds 70 percent If the percent utilization exceeds 70 percent the bar on the strip chart changes from black to red for the period of time during which the threshold is exceeded To set a threshold value 1 Click MB1 anywhere in the instrument to be modified The instrument appears to be depressed to show that it has been selected 2 Press and hold MB3 to display the pop up editing menu The menu lets you edit the following items m Ranges and thresholds m Patterns and colors 3 Choose Ranges and Thresholds by moving the cursor over this selection and releasing the MB3 button For example the following dialog box is displayed when one of the graphs on the System Overview panel is selected as shown in the following screen Wl Automatic Scaling Maximum Data Valu 100 00 inimum Data Value o Number of Time Units 30 C Low Threshold 20 00 L High Threshold 90 00 This dialog box is also used to set the scale for an instrument The maximum and minimum data values to be displayed for the bar or strip chart can be set In addition the strip chart can be set to automatic scaling whereby the height of the chart is dynamically adjusted to match the largest data value shown 338 Performance Manager Administrator Guide Change the Colors and Patterns The Ranges and Thresholds dialog box lets you set two levels
64. CSV File to a Windows Machine 22222020 000 e 562 Step 5 Create the Graph in Excel 2222 220 llla 562 Glossary 565 Index 577 Contents 11 Chapter 1 Introduction This chapter introduces CA Performance Management for OpenVMS and discusses its features supported configurations and system requirements This section contains the following topics CA Performance Management for OpenVMS see page 13 Performance Manager see page 14 Performance Manager Features see page 15 DECwindows Interface see page 17 What to Expect from Performance Manager see page 18 CA Performance Management for OpenVMS CA Performance Management for OpenVMS includes layered products designed to reduce the time and effort required to manage and monitor system performance and to plan for future resource requirements These products can be used with standalone and with OpenVMS cluster systems CA Performance Management for OpenVMS includes the following products m Performance Manager m Performance Agent wm Accounting Chargeback Performance Manager and Performance Agent share a common database and basic set of utilities Any one component may provide these utilities on behalf of the other components of the same version as shown in the following items m Performance Agent gathers manages classifies and archives OpenVMS system data It provides the following functions Data collection and storage Data archiving Dump repor
65. Chapter 6 Performance Manager Commands 199 ADVISE PERFORMANCE GRAPH This command produced two of three graphs because threshold quantity for the last graph was lowered ADVISE PERFORMANCE GRAPH BEGINNING 10 ENDING 11 NODE YQUEM _ TYPE CPU U THRESHOLD 25 CPU M THRESHOLD 35 TOP CPU I THRESHOLD 15 SELECT GREATER 90 0UTPUT REGIS The previous command generated none of the graphs because in all cases 90 percent of the graph points did not exceed the specified thresholds STACK NOSTACK Stacks the values for each category on the graph Use NOSTACK to overlay the values on the graph ReGIS graphs using NOSTACK may cause some occlusion if you do not specify FORMAT ReGIS CHARACTERISTICS LINE also If you are requesting a series of graphs in one command you can override the NO STACK qualifier by specifying the NO STACK keyword following each graph type See Chapter 4 for an illustration of the use of the NOSTACK qualifier and for additional information about default behavior TYPE NO graph_type NO STACK Y_AXIS_MAXIMUM n TITLE string ALL_GRAPHS NO STACK THRESHOLD m Y AXIS MAXIMUM n CUSTOM see below TYPE CUSTOM Specifies which of the graphs you want generated Use the TITLE keyword to override the Performance Manager supplied title The text string may be a maximum of 40 characters The STACK keyword for a particular graph type overrides the setting established by the
66. Chapter 8 Use the DECwindows Motif Interface 315 How You Customize History File Descriptors To create modify or delete history file descriptors in the parameters file 1 Pull down the Parameter submenu and choose the History File Descriptors menu item The Performance Manager displays the History File Descriptors dialog box JT HIS MONTHLY MODEL MONTHLY USER History Descriptor Data Reduction Scheme Archive Schedule Legend Granularity E Collection On MA Collection Off Periodicity None Press ME3 Inside clock zm 2 to copy Schedule Interval 72 Delete files after days Workload Classification o Model Data unlimited Classify by families Excluded Workload Families Included Workload Families ABC FAM ALL USERS ANNS IMAGE TRANS ANNS TERM TRANS KI CONSOLIDATED Clear Modify rese The menu bar contains Control and Help menus A list of history file descriptors appears at the top of the dialog box 2 To close the dialog box pull down the Control menu and click the Exit menu item 316 Performance Manager Administrator Guide How You Customize The History File descriptor dialog box lets you do the following actions Create a history file descriptor Delete a history file descriptor Modify a history file descriptor For a detailed description of history file descriptors see the Performance Agent Ad
67. DSK FIL LOC PRO TAP QUOTA CACHE AR Derived Attempt rate per second to the quota cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP QUOTA CACHE HR Derived Hit ratio to the quota cache for the local node for the current interval record Calculated by dividing the number of quota cache hits by the number of quota cache attempts hits misses times 100 Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP RDTS IN LISTMET F RDT MAX This contains the total number entries in the I O request descriptor table for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 495 Performance Manager Data Cells RDT WAIT RATEMET F RDT QUE This contains the number entries in the I O request descriptor table in a wait queue for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SCS ADAPTERNAMECFG A ADAPTER This contains the adapter name string of the remote node s adapter for the current configuration subrecord e g CIXCD Data Type INDEX Domains CFG SCS ADAPTER IDCFG L ADAPTER ID This contains the adapter code of the remote node s adapter for the current configuration subrecord See PSPASLIB for a list of known codes Data Type NUMERIC Domains CFG SCS NODENAMECFG A NODENAME This contains the
68. Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES IN FPGMET F FPG This contains a value representing the average number of processes in free page wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES_IN_HIBMET_F_HIB This contains a value representing the average number of processes in hibernate wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 478 Performance Manager Administrator Guide Performance Manager Data Cells PROCESSES IN HIBOMET F HIBO This contains a value representing the average number of processes in outswapped hibernate wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES_IN_LEFMET_F_LEF This contains a value representing the average number of processes in local event flag wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES IN LEFOMET F LEFO This contains a value representing the average number of processes in outswapped local event flag wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCE
69. Domains CFG TOP_DIRIO_PROCESS_DIRIO Derived This contains a value representing the highest direct I O operations rate for a process subrecord on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 523 Performance Manager Data Cells TOP DIRIO PROCESS X Derived This contains an index pointing to the process subrecord that has the highest direct I O operations rate on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP_DIRIO_PROC_TOPDSK_X Derived This contains an index pointing to the disk record with the highest operations rate for the process subrecord that has the highest direct I O operations rate on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK TOP DISKS PROCESS X Derived This contains an index pointing to the process subrecord that has the highest I O operations rate to the highest I O rate disk on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP_DSKIO_PROCESS_DSKIO Derived This contains a value representing the highest disk I O operations rate for a process subrecord on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TOP_D
70. Domains COM CFG CPU DSK FIL LOC PRO TAP BLKS_FREE_IN_NONPAGED_POOLMET_F_NP_FREE_BLOCKS This contains the number of free blocks in non paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BLKS_FREE_IN_PAGED_POOLMET_F_PG_FREE_BLOCKS This contains the number of free blocks in paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BLOCK_REQUEST_DATAS_INITIATEDSCS_F_REQDATS This contains the value representing the number of block transfers initiated per second for request data s on the local node to the remote node for the current configuration record and interval Data Type NUMERIC Domains CFG 416 Performance Manager Administrator Guide Performance Manager Data Cells BLOCK REQUEST DATAS INIT TALLY Derived This contains the sum of the values representing the number of block transfers initiated per second for request data s on the local node to the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG BLOCK_SEND_DATAS_INITIATEDSCS_F_SNDATS This contains the value representing the number of block transfers initiated per second on the local node to the remote node for the current configuration record and interval Data Type NUMERIC
71. FOCU CPU OUT EXAMPLE 3 CPU INFO The following statements describe the columns in the previous example m Primary Secondary Key Lists the primary and secondary key identifiers for the process detail lines The keys may be Mode Username Imagename Processname Account UIC group Workloadname or PID m CPU Utilizatn Provides the amount of CPU time consumed by the process or processes represented as a percentage of the node s total capacity If this is a cluster report PROCESS CLUSTER then the figure is a percentage of the cluster s total capacity The process CPU time is scaled by the node s VUP rating power rating m Cpu time Min Provides the amount of CPU time consumed by the process or processes in minutes wm Uptime Min Provides the total amount of wall clock time during which the process es were resident on the system in minutes m Response tim sec Represents the average response time per terminal input operation for the process or processes m Terminal Inpt Sec Provides the average number of Terminal Inputs completion of a read QIO to the terminal per second for the process or processes m Terminl Chrs sec Provides the average number of bytes characters per second transferred to or received from the terminal for the process or processes Chapter 3 Evaluate Performance in Detail 45 Performance Evaluation Report m Pri Shows the process s current priority and base priority For more than one proc
72. Formatted Pie Chart Output illustrates CSV formatted output for a pie chart It was generated with the following command ADVISE PERFORMANCE PIE CHART BEGINNING 22 MAR 2006 12 00 _ ENDING 22 MAR 2006 14 00 NODE BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY FORMAT CSV _ TYPE CUSTOM SYSTEM METRIC CPU TOTAL COMPOSITE PSPA CUSTOM GRAPH Nodes BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY Date 22 MAR 2006 12 00 14 00 The whole pie represents 11 79 CPU Total Busy Item Value Percent of Total BUGDEV 5 68626 48 216 BYOB 3 59973 30 523 CINAMN 0 96963 8 222 ORIPAS 0 49438 4 192 BATCH 0 49370 4 186 CHATTY 0 36808 3 121 MIGHTB 0 18164 1 540 Data Resolution with X POINTS For ReGIS and PostScript graphs the X POINTS keyword indicates the number of data points to plot along the x axis of the graph and is specified with the FORMAT REGIS X_POINTS number qualifier from a DCL command or X_POINTS number qualifier from command mode For Tabular graphs the X POINTS keyword specifies the number of data points to present in the report and is specified with the FORMAT TABULAR X_POINTS number qualifier The valid range for X POINTS is 2 to 480 the default is generally from 45 to 90 but is computed to produce an even time interval per point The following are examples of specifying the X POINTS keyword DCL command ADVISE PERFORMANCE GRAPH FORMAT REGI
73. I O statistic for each node is that node s contribution to the total cluster I O load Chapter 3 Evaluate Performance in Detail 103 Tabular Report Sections In the case of a configuration with one node with direct access to the disk and any number of nodes with remote access I O statistics for the node with direct access includes I Os from any nodes with remote access to the disk Therefore to compute the direct access node s contribution the software subtracts the number of I Os of the nodes with remote access from the I O information of the node with direct access In the case of two nodes with direct access to a disk with any number of nodes with indirect access both direct nodes I O statistics include I O from any remote nodes Therefore it is not possible to compute the By Node contribution to the cluster total of the nodes with direct access to the disk The By Node contribution of the nodes that access the disk remotely is each node s I O statistics For example the software would calculate By Node rates for disks DU1 and DB2 in the example configuration as follows For disk DU1 with the HSC all I Os come to this disk through host direct access Nodes A and B The I O information for these nodes includes I O from the served remote access Node C There is no way to distinguish whether Node C s I Os were processed by Node A or Node B Therefore calculating the By Node contribution for Nodes A and B which directly access the d
74. IIIIIIIIIIIIIIIIIIIIIIIITIN IB IIIIIIIIIIIIIIIIIIIIIIIITIN II IIIIIIIIIIIIIIIIIIIIIIIITIB II IIIIIIIIIIIIIIIIIIIIIIIIIIBNII IIIIIIIIIIIIIIIIIIIIIIIIIIBNII IIIIIIIIIIIIIIIIIIIIIIIIIIBBII IIIIIIIIIIIIIIIIIIIIIIIIIIIBII N IIIDIIIIIIIIIIIIIIIIIIIIIIIIITIIBIIN N IN INDIIDIIIIIIIIIIIIIIIIIIIIIIIIIIIBIII N DI IIDIIDIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIN IN DI IDDIIDIIIIIIIIIIIIIIIIIIIIIIIIIIIIITIN II DDIDDDIIDIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIINIIN DDIDDDDIDIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIN DDIDDDDIDIIIIIIIIIIIIIIIIDIIIIIIIIIIIIIIIIIINN DDIDDDDIDDIIIIDIDDIIIIIIDDDIIIIIIIDDIIIIIIIIINNN 4 N DIDDDDDDDDXDXDDDXDDDDDDIIIIDDDDDDIIIIXXDDDDDDDDDDDDDIIIDIIIININNNNNNDN DDDDDDDDDXXXXDDDXXXXDDDDDDDDDXXDDDDDDXXXXDDXXDDXXDDDDDIDDDDIIDDNDDIIIDN DDDXXXXDXXXXXXXXXXXXXXDXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXDDXDDDDDXXDDDDDDDD 13 00 13 20 13 40 14 00 14 20 14 41 15 01 15 21 15 41 16 01 16 22 16 42 Min 12 72 Max 100 00 Average 62 51 120 Performance Manager Administrator Guide Format Graphs and Pie Charts The following table Tabular Formatted Output illustrates tabular formatted output Tabular output is in 132 column format It was generated with the following command ADVISE PERFORMANCE GRAPH BEGINNING 19 FEB 2006 13 00 ENDING 19 FEB 2006 17 00 TYPE CPU UTILIZATION NODE GALLO OUTPUT CPU TAB LIS FORMAT TABULAR CPU UTILIZATION Node GALLO Date 19 FEB 2006 13 00 17 00 Metric Values are Stacked Added to the left Units Pe
75. INVESTIGATE SYSTEM OVERVIEW DISPLAY INVESTIGATE MEMORY DISPLAY INVESTIGATE IO DISPLAY INVESTIGATE CPU DISPLAY INVESTIGATE LOAD BALANCE DISPLAY INVESTIGATE IO MEMORY PAGEFILE PID n RESOURCE CPU DISPLAY RESOURCE DISK DISPLAY Control the Displays Function Provides a composite display for all disks servers or volumes Provides a multi node display for the desired metric Stops the input data stream Advances to the next interval while display is suspended Provides a process display for the specified image name Provides the ReGIS display for investigating the selected system Provides the ReGIS display for investigating the selected system s memory Provides the ReGIS display for investigating the selected system s IO rate Provides the ReGIS display for investigating the selected system s CPU Provides the ReGIS or ANSI display for investigating the selected system s load balance Provides the ReGIS display for investigating the last selected display or the system display if none had been selected Provides a multi node display of direct paging and swapping IO rate Provides a multi node display of memory utilization Provides a multi node display of Pagefile Utilization Provides a process display for the specified PID Provides a multi node display of CPU resources Provides a multi node display of disk resources Chapter 11 Use the Character Cell Real time D
76. INVESTIGATE Keypad The following illustration shows the functions of the INVESTIGATE keypad Resource Display KEYPAD DEFINITION for the INVESTIGATE DISPLAY Next and Prev Screen Allow you to view the various INVESTIGATE displays SYS CPU MEM IO and LOAD Switch to Z Exit PSPA RT the Single Node Display Evaluate Performance Using the Investigate Displays The System and Load Balance displays provide graphic information to determine which of the three main system resources memory I O or CPU is a limitation To investigate a limitation in more detail use the Memory I O and CPU displays Note For a detailed description of a system tuning methodology see HP s OpenVMS Performance Management guide 392 Performance Manager Administrator Guide Evaluate Performance Using the Investigate Displays When determining a limitation in a main resource investigate the resources in the following order 1 Memory 2 I O 3 CPU The order is important because memory limitations cause paging and swapping which lead to I O and CPU problems Begin your tuning investigation by using the System Overview or Load Balance displays as shown in the following screens File tdi Commands Options Print GALLO VEILSEEEELEISEE ERES ESSE SEE ESEEEE EL S JF fe Edit Commands Options Print GALLO LS FEB L99D 14 56 It Chapter 11 Use the Character Cell Real time Display 393 Evaluate Performance Using the I
77. LOC PRO TAP XQP VOL AND DIR SYNCH LOCK RATEMET F SYNCHLCK Average XQP directory and volume synchronization lock rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP VOL SYNCH LOCK RATEMET F VOLLCK Average XQP volume synchronization lock rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP VOL SYNCH LOCK WAIT RATEMET F VOLWAIT Average XQP volume synchronization lock wait rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 533 Appendix D Estimate Virtual Memory Needs This appendix helps you to estimate the virtual memory requirements for the Performance Manager when performing the following tasks m Generating Reports and Graphs using the DCL user interface m Selecting daily or history data using the Command Mode interface m Selecting daily or history data using the DECwindows interface The virtual address space that a process may use is governed by smallest of either the process quota PGFLQUOTA or the SYSGEN Parameter VIRTUALPAGECNT Since an image is generally forced to exit when its virtual address space is exhausted it is best to plan to have a sufficient amount for the desired task This section contains the following topics How P
78. Manager Data Cells 455 Performance Manager Data Cells Domains CLU HSC TYPE HSC95 Derived This contains a Boolean value 1 if the current HSC controller is an HSC95 Data Type NUMERIC Domains CLU IDLEMET F IDLE Average percentage of CPU idle time for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IDLE PROC WITH BIG WS Derived This contains a Boolean value representing the presence of one or more idle processes with overly large working sets on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IMAGE ACTIVATION RATEMET F IMGACTS This contains the number of image activations per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IMAGE HUNG IN MWAIT NOT RWAST Derived This contains a Boolean value of zero or one one true representing the presence of an image hung in an MWAIT state other than an AST resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 456 Performance Manager Administrator Guide Performance Manager Data Cells IMAGE HUNG IN RWAST Derived This contains a Boolean value of zero or one one true representing the presence of an image hung in an RWAST resource wait state on the local node for the current interval Data Type NUMERIC Domai
79. Manager Data Cells 529 Performance Manager Data Cells WORKING SET FAULT TALLY Derived This contains the sum of the values representing the soft page fault rate per CPU second for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO WORKING_SET_GLOBAL_PGSPRO_F_GPGCNT This contains a value representing the number of physical global pages in the current working set of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO WORKING_SET_LISTPRO_F_WSSIZE This contains a value representing the number of pages allowed in the current working set of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO WORKING_SET_LIST_TALLY Derived This contains the sum of the values representing the number of total pages in the working sets for all of the current process subrecords which were selected by the most recent PROCESS_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO WORKING_SET_PRIVATE_PGSPRO_F_PPGCNT This contains a value representing the number of physical private pages in the current working set of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 530 Performance Manager Ad
80. Manager provides a Disks display when you enter the following command PSRT5 DISKS The following illustration is an example of a Disks display File Edit Commands Options Print Yolure I O Rate i 1i nodes T 102 OOVER WESTON BOLTON toston NECOMAM WALTHAM Press PF2 to view the keypad for the Disks display The following illustration shows the functions of the Disks display keypad Resource Invest KEYPAD DEFINITION Display igate Display for the DISKS DISPLAY Zoom Out to aZ Exit PSPA RT Multi Node Display Press KP6 to scroll through Disk keys and KP9 to scroll through disk metrics To return to the multi node display press KPO 384 Performance Manager Administrator Guide Display Rules Information Display Rules Information To obtain the Rules display m Enter the RULES command If any factory or user rules fire for the last data record processed the rule ID and a brief explanation are displayed If data is displayed at less than two minute intervals then only user rules are displayed Note Rules in the Cluster and Summary domain are not displayed The following illustration is an example of a Rules display vaut File Edit Commands Options Print Help I Rule s VQUEM at time 15 20 00 COOTO CPU bottleneck reduce demand or add CPU power Display RESOURCE Information The purpose of the Resource displays is to permit evaluation of resource utilization in interactive
81. OCT 00 00 00 to 1 OCT 12 16 44 An N indicates NO DATA The following statements are keyed to the CPU Utilization histogram example 1 Explanatory list of symbols in the histogram columns 2 Percentage of CPU time used by categories of processes and system overhead 3 Number of hours spanned by the reporting period This histogram reflects a reporting period of 12 hours Chapter 3 Evaluate Performance in Detail 65 Histograms Physical Memory Use Histogram The following example shows a physical memory use histogram Histograms DEMAND VAX 8700 Page 6 PA Vx x Tuesday 06 SEP 2005 00 00 to 23 59 Legend 1 Modified List Free List User Ws Tot Wss For All System Users PHYSICAL MEMORY USAGE VMS Allocated of memory 2 95 90 85 80 75 70 65 60 55 4 50 45 40 35 u 30 VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV 20 15 VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV 5 3 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 25 10 SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV VVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVVV 3 Each Column represents approximately 20 minutes starting fr
82. PROCESS commands or the calls to the CREPRC System Service or LIBSSPAWN Runtime Library Routine As a last resort for detached processes increase the PQL DWSEXTENT SYSGEN parameter if it is responsible for establishing the working set extent value for the detached process Total number of users needing an increase User name s CORREY CONDITIONS 1 SOFT FAULT RATE GE 100 00 SOFT FAULT SCALING OR HARD FAULT RATE GE 10 00 HARD FAULT SCALING 2 IMAGE ACTIVATION RATE LT 0 50 IMG ACT RATE SCALING 3 SYSGEN MPW HILIMIT SYSGEN FREEGOAL LT MEMORY PAGES NOT ALLOC TO VMS 0 04 0 05 0R PAGES ON FREELIST GE SYSGEN FREEGOAL OR HARD FAULT RATE GE 10 00 EVIDENCE User CORREY HARD FAULT SCALING 4 SYSGEN MPW HILIMIT SYSGEN FREEGOAL GE MEMORY PAGES NOT ALLOC TO VMS 0 05 OR SOFT FAULT RATE GE 100 00 SOFT FAULT SCALING 5 PROCESSES NEED MORE EXTENT EQ 1 00 6 OCCURRENCES GE 1 Working Set Time range Avg Image Faults sz No batch jobs Image W Set flts of Free of are flagged Name Size cpusc Hrd Soft List Tims 4 6 200 11 34 13 34 LOGINOUT 11 34 15 14 TEST31 11 34 13 30 SET 2048 4000 batch 13 16 13 18 TEST33 Chapter 2 Analyze Performance 23 Analysis Reports The following statements are keyed to the columns in the previous Memory Rule report 1 User name associated with a process 2 Working set quota and working set extent 3 Time range for which select
83. Performance YQUEM VAX 6000 440 Evaluation Tuesday 26 JAN 1997 09 00 to 10 00 Primary Secondary Key Img Cpu time DirI O BufI O SftFlt HrdFlt Ave Max Ave WS WS Pid Image Pri State Cnt Min Sec Sec Sec Sec WSsize WSsize Global Default Quota Extent 29400201 Swapper 29400206 CONFIGURE 29400209 ERRFMT 294004CO dcl a j 452 474 CMS 133 a P 509 862 DIRECTORY 005 P x r 477 477 MAIL 046 x F 1115 1317 29400A41 dcl 022 r x 379 554 DELETE CUR 002 2 E 308 308 DIRECTORY CUR 002 a 495 495 SEARCH CUR 006 x 325 325 SHOW CUR 002 E 375 375 TPU CUR 164 I 1167 1497 User Command ADVISE PERF REPORT PERF NODE YQUEM BEG 26 JAN 1997 09 00 00 00 END 26 JAN 1997 10 00 00 00 INCLUDE PROC PROC FOCU GEN OUT EXAMPLE 3 GEN The following statements describe the columns in the previous report m Primary Secondary Key This column lists the primary and secondary key identifiers for the process detail lines The keys may be Mode Username Imagename Processname Account UIC group Workloadname or PID m Pri This column shows the process s current priority and base priority For more than one process they are averages weighted by the process s uptime m State This column represents the scheduling state of the process for the most recent data For more than one process this column reflects the state of only one of these processes Img Cnt This column provides the number of im
84. Q PostScript Btw Q Tabular O csv Chapter 8 Use the DECwindows Motif Interface 287 How You Display Analyzed Data The Save Graph or Save Pie dialog boxes allow you to do the following actions m Specify a file name m Select an output format More output formats are available for graphs than for pie charts m Specify the title 2 Click OK to save the graph or pie chart to a file OR Click Cancel to remove the window from the screen with no data being saved Edit the Graph Format Pull down the Edit menu You can change the following aspects of the graph Stack Specifies that the values for each category on the graph are to be stacked This is the default Graph Window File Edit Display Display Top Display Custom TOP CPU IMAGES Release on the button to turn Stack off The Performance Manager software displays a line mode graph to avoid occlusion The values for each category are overlaid or unstacked If you request an unstacked graph of top metrics the items grouped together are displayed as an average instead of a total 288 Performance Manager Administrator Guide How You Display Analyzed Data Pie Chart Specifies that the display is to be a pie chart A pie chart produces a slice for each item using size to depict average units of measure over the specified time period When pie charts are displayed stack and line mode toggle buttons do not affect the display For more information see the cha
85. QUARTERLY Causes graphs to depict a summarization of a specified time period The selected data is averaged into the time period selected See Chapter 4 for more information If you also use the SCHEDULE or DATES qualifiers the DAILY and WEEKLY graphs are trimmed to show only the selected hours If history data with the periodicity attribute is selected the AVERAGE value is automatically set to that periodicity value This is true regardless of whether the AVERAGE qualifier is used Chapter 6 Performance Manager Commands 191 ADVISE PERFORMANCE GRAPH BEGINNING date Specifies the beginning date and time of data selected for graphing Where date represents the date and time in standard DCL format The date and time format is the standard DCL format either absolute or relative If you do not specify the BEGINNING qualifier the Performance Manager uses 00 00 00 on the same day for which the ending date and time is specified If you do not specify an ENDING qualifier the Performance Manager uses 00 00 00 of the current day as the default beginning time You can also use the keywords TODAY and YESTERDAY See HP s OpenVMS User s Manual or access the HELP topic SPECIFY DATE TIME for complete information on specifying time values BEGINNING is incompatible with the DATES qualifier COLLECTION DEFINITION collection definition name Specifies the name of the Collection Definition and hence the collected data that you desi
86. Rules file Rules file constructs Saturation Decimal values Literal symbols Tally data cells Numeric data cells Boolean data cells Scan routine data cells String operators Numeric binary operators Parentheses for precedence After all the data has been processed when creating a report the Performance Manager examines the number of rule occurrences for each rule If the rule occurrence threshold is met for a particular rule the rule is said to fire For each rule that fires an entry is made in the Analysis report The entry may include evidence and conclusions A rule element A five character alphanumeric code enclosed in braces for example M0010 which uniquely identifies a rule A zero for the second character is reserved for Digital use only Each time all the rule conditions for a given rule are true there is one rule occurrence See Rule firing The rules compiler generates a binary knowledge base file name KB from an ASCII rules source file name VPR The rules file is a compiled knowledge base file name KB A Performance Manager rules source file name VPR The following five constructs can exist in a Rules File Comment construct Disable construct Literal construct Rule construct Threshold construct Saturation is the point at which response time at a device becomes substantially higher than the service time Glossary 573 Schedule file Shadow set Soft page fault Spli
87. SUMMARY IS AN ALPHA Derived This contains a one if the hardware model is an Alpha zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP IS AN IA64 Derived This contains a one if the hardware model is an Integrity server zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP IS A VAX Derived This contains a one if the hardware model is a VAX zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 460 Performance Manager Administrator Guide Performance Manager Data Cells KB MAPPEDSCS F KBYTMAPD This contains the value representing the number of kilobytes mapped per second between the local node and the remote node indicated by the cell SCS NODENAME for the current configuration record and interval Data Type NUMERIC Domains CFG KB MAPPED TALLY Derived This contains the sum of the values representing the number of kilobytes transferred per second between the local node and the remote node for all the current configuration s
88. Specify a list of strings separated by commas spaces or tabs to generate reports and graphs for all process records with the processname matching any of the specified strings The match string is case sensitive so if the process names have any lower case letters spaces or tabs use double quotes when you enter the value for example RTserver Accountnames Specify a list of strings separated by commas spaces or tabs to generate reports and graphs for all process records with the accountname matching any of the specified strings Chapter 8 Use the DECwindows Motif Interface 273 How You Select Data for Analysis Workloadnames Specify a list of strings separated by commas spaces or tabs to generate reports and graphs for all process records associated with any of the specified workloads This filter is valid only if the Classified By option is used to specify a classification scheme for your workload data Uics Specify a list of UICs separated by commas spaces or tabs to generate reports and graphs for all process records with the UIC matching any of the specified UICs An asterisk may be used to wildcard either the group or user field of the specified UICs Pids Specify a list of PIDs separated by commas spaces or tabs to generate reports and graphs for all process records with the PID matching any of the specified PIDs Volumenames Specify a list of strings separated by commas spaces or tabs to genera
89. TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU 426 Performance Manager Administrator Guide Performance Manager Data Cells CPU MP SYNCHCPU F MP SYNCH This contains the value representing the percent of time spent in MP SYNCH mode for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU MP SYNCH TALLY Derived This contains the sum of the values representing the percent of time in MP SYNCH mode for the current CPU subrecords which were selected by the most recent CPU SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU CPU ONLYMET F SPMCPUONLY Percentage of time that a CPU was busy and all disk devices were idle for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CPU PHYSICAL ID Derived This contains a string indicating the CPU s physical ID of the current CPU subrecord for the current interval Data Type STRING Domains CPU CPU SCAN Derived Provides the count of CPU subrecords for which the specified rule condition is true The condition will be evaluated for each CPU subrecord Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU Appendix C Performance Manager Data Cells 427 Performance Manager Data Cells CPU SUPERCPU F SUPER This contains the value representing the percent of t
90. TALLY WORKING SET LIST TALLY WORKING SET DEFAULT TALLY WORKING SET QUOTA TALLY WORKING SET EXTENT TALLY PROCESS UPTIME TALLY PROCESS IMAGE ACTS TALLY PROCESS COM PERCENT TALLY PROCESS DISK THRUPUT TALLY PROCESS DISK IO TALLY TAPE IO TALLY TAPE ERROR TALLY Index specifiers are data cells that indicate a specific occurrence of a sub record that has a unique characteristic in one of these domains COMMUNICATION m CONFIGURATION m CPU m DISK m FILE m PROCESS m TAPE During analysis the Performance Manager reads an interval record The Performance Manager evaluates rules in the following domains m LOCAL m PROCESS m DISK m FILE m CPU 162 Performance Manager Administrator Guide Implement Changes COMMUNICATION CONFIGURATION m TAPE Data cells available in the LOCAL domain are available to rules in all of these domains You can reference data cells in the PROCESS DISK FILE CPU COMMUNICATION CONFIGURATION or TAPE domains and not in the LOCAL domain directly by rules within that domain or indirectly with an index specifier to data cells in any of the other domains Each index specifier data cell has a target domain The target domain indicates the name of the domain of the desired data cell For example the index specifier TOP_CPU_PROC_X points to a specific process sub record for the current interval You might use the index specifier in rule definitions in the LOCAL DISK FILE CPU COMMUNICATION
91. THRESHOLD has this value if the current disk sub record is of this type Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Chapter 5 Customize the Knowledge Base 141 Components of Rules 142 Performance Manager Administrator Guide Name TD T15 RM05 TD T20 RA80 TD T21 RA81 TD T22 RA60 TD T25 RD51 TD T27 RD52 TD T28 RD53 TD T30 RA82 TD T31 RD31 TD T32 RD54 TD T34 RRD50 TD T36 RX33 TD T37 RX18 TD T38 RA70 TD T39 RA90 TD T40 RD32 TD T42 RX35 TD T43 RF30 TD T44 RF71 TD T45 RD33 TD T46 ESE20 TD T47 TU56 TD T48 RZ22 TD T49 RZ23 TD T50 RZ24 TD T51 RZ55 TD T52 RRD40S TD T53 RRD40 TD T54 GENERIC DK TD T55 RX23 Default Value 26 27 28 20 10 20 26 33 18 26 20 10 10 32 34 20 10 24 24 18 240 30 24 24 32 31 20 20 30 10 Description Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous C
92. The metric graphed is CPUTIME TYPE TOP CPU WORKLOADS Top CPU Workloads Plots the top five plus Other workloads as consumers of CPU time The metric graphed is CPUTIME TYPE TOP CR W Top Credit Wait Rate Plots five nodes with the highest rate of credit waits plus Other resulting when a connection has to wait for a send credit The metric graphed is CR W TYPE TOP DGS D Top Datagrams Discarded Plots five nodes with the most datagrams discarded plus Other resulting when application datagrams are discarded by the port driver The metric graphed is DGS D TYPE TOP_DGS_R Top Datagrams Received Plots five nodes with the most datagrams received plus Other resulting when the local system receives datagrams over the connection from the remote system and given to SYSAP The metric graphed is DGS_R TYPE TOP DGS S Top Datagrams Sent Plots five nodes with the most datagrams sent plus Other resulting when application datagrams are sent over the connection The metric graphed is DGS S Chapter 6 Performance Manager Commands 221 ADVISE PERFORMANCE GRAPH TYPE TOP DIRIO IMAGES Top Direct I O Images Plots the top five plus Other Images creators of direct I O by image name The metric graphed is DIRIO TYPE TOP DIRIO USERS Top Direct I O Users Plots the top five plus Other Users creators of direct I O by user name The metric graphed is DIRIO TYPE TOP DIRIO WORKLOA
93. To customize the Process Statistics Report pull down the Customize menu and release on the menu item you want The Customize menu lets you enter any of the following Cluster stats by node Release on the Cluster stats by node menu item to turn on or off the reporting of cluster wide process statistics by node Report keys To re sort the process data release on the Report keys menu item Performance Manager displays the Primary and Secondary Keys dialog box The Primary and Secondary keys you specify are applied when you use the SAVE AS option to save process statistics as shown in the following screen Control Buttons m The OK button removes the dialog box sorts the process data and re displays the report m The Apply button sorts the process data and re displays the report m The Cancel button removes the dialog box without changing the keys Chapter 8 Use the DECwindows Motif Interface 283 How You Display Analyzed Data Tabular Report Sections Release on the Tabular report menu item from the Display menu to open the Performance Manager Tabular Report window as shown in the following screen File Nade interval Data Cluster Interval Data Node Config Section _ xm H d EE revo To navigate through the Tabular Report click the arrow buttons on either side of the Configuration label This updates the window with the Summary Process Exten
94. Types Brief Descriptions Boolean 1 0 True or 0 False Numeric Floating point value String An ASCII string Time Clock time for data Scan routines Loads tally data cells and returns sub record count Tally Derived data Index specifier Index to a sub record Boolean Data Cell A Boolean data cell is a value provided by the Performance Manager software that represents the result of applying a commonly needed condition to a domain or a subset of internal records A Boolean data cell has a value of true 1 0 or false 0 For example the ANY DISK OVER THRESHOLD data cell is set to either true or false If it is set to TRUE at least one disk for the current Performance Manager interval record has exceeded the I O rate threshold for the given disk type If this Boolean operator did not exist you would have to use an expression similar to the following code DISK SCAN DISK IO RATE GE DISK IO RATE THRESHOLD GT 0 158 Performance Manager Administrator Guide Numeric Data Cell String Data Cell Time Data Cell Data Cell Types and Use A numeric data cell provides a floating point value For example NUMBER OF PROCESS is a numeric data cell in the LOCAL domain that contains the floating point value for the average number of processes resident on the system during the given two minute period for the current node A string data cell provides the actual string when the cell is used as evidence However when you use it in a
95. Utilization REPORT The REPORT command generates Performance Manager Analysis Reports Performance Evaluation Reports Tabular Reports and Histograms Format REPORT report_keyword Chapter 7 Use Command Mode Commands 249 REPORT Description Use the REPORT command to produce Performance Manager reports The Performance Manager can generate reports using either daily or historical data The following table lists all the REPORT command options Options Option Function ANALYSIS Displays the full Analysis Report BRIEF ANALYSIS Displays the brief Analysis Report PERFORMANCE EVALUATION Displays the Performance Evaluation Report HISTOGRAMS Displays standard ANSI graphs of CPU memory and I O use The Image Residence histogram is not available in Command Mode TABULAR FINAL BYCLUSTER BY Displays an overview of the system NODE activity on a per node basis or cluster wide Subsections of this report can be selected or omitted with the SECTION qualifier Tabular process statistics require that you select PROCESS ALL to save by PID details for each process See also SELECT Qualifiers EXPLAIN NOEXPLAIN Specifies for the Full Analysis Report whether to include the rule s conditions and evidence in the report output By default the Performance Manager uses the EXPLAIN qualifier for a batch process or if OUTPUT is specified or asks you if the process is interactive HOTFILE_LIMIT n Specifies the maximum n
96. a subinterval specified during data selection The menu entries Node Interval Data and Cluster Interval Data provide access to the subinterval statistics windows If the Tabular Interval option was not chosen during data selection these menu options are desensitized The node statistics classes include memory CPU paging IO XQP lock SCS and process statistics Each class appears in a separate window If the node selected in the parent window is changed the interval statistics windows are all updated with that node s data The beginning time stamp of each interval is included in the display The cluster statistics classes include memory CPU and lock statistics either in terms of minimum maximum average displays or by node displays For details on the data displayed in the different report sections see the chapter Evaluate Performance in Detail see page 33 Chapter 8 Use the DECwindows Motif Interface 285 How You Display Analyzed Data Graphs To view graphs or pie charts of selected data release on the graph item of the Display menu The Performance Manager Graph Window is displayed as shown in the following screen m Graph Window File Edit Display Display Top Display Custom CPU UTILIZATION Each graph has the same basic format The components are these Title The title is centered at the top of the graph and identifies the type of graph Subtitle The subtitle gives the node name or list of node names f
97. a Boolean value where a 1 means the process logged in For history data it means one or more processes This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 484 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS IMAGE LOGOUTPRO B LOGOUT This contains a Boolean value where a 1 means the process logged out For history data it means one or more processes This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS IMAGE TERMINATIONPRO B IMGTRM This contains a Boolean value where for CPD data a 1 means the process s terminated the image which triggered the creation of this subrecord and for History data a 1 means that one or more image terminations took place for data summarized into this process subrecord This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS RESPONSE WAITPRO F RESPONSE WAIT This contains a value representing the number of milliseconds elapsed from the completion of the most recent terminal read to the end of the current interval represented by this record Data Type NUMERIC Domains PRO PROCESS SCAN Derived Provides the count of process subrecords for which the specified rule expression is true on the local node for the current inter
98. a much higher rate than shown in the Performance Manager Thresholds table then you might want to increase your disk operations rate thresholds to account for this performance and eliminate these extraneous rule firings Suggested numbers for these scenarios are as follows m If you have an older RF31 as opposed to the newer RF31T then you should change the threshold TD T56 RF31 to 34 m If your workloads on any of the following disks benefit from disk caching then you might want to set the following thresholds TD T43 RF30 31 TD T48 RZ22 40 TD 144 RF71 31 TD T49 RZ23 41 TD T56 RF31 96 TD T50 RZ24 51 TD T57 RF72 55 TD T51 RZ55 54 TD T75 RFH31 96 TD T59 RZ25 65 est 168 Performance Manager Administrator Guide Implement Changes TD T76 RFH72 55 TD T60 RZ56 58 TD T77 RF73 63 TD T61 RZ57 62 TD T78 RFH73 63 TD T66 RZ23L 47 TD T81 RF35 87 TD T68 RZ571 62 TD T82 RFH35 87 TD T70 RZ58 70 est TD T83 RF31F 56 TD T84 RZ72 63 est For the older RF31 TD T85 RZ73 63 TD T56 RF31 51 TD T86 RZ35 87 TD T75 RFH31 51 TD T87 RZ24L 55 est TD T88 RZ25L 62 est With HSC caching TD T89 RZ55L 59 est TD T80 RA71 97 TD T90 RZ56L 60 est TD T79 RA72 97 TD T91 RZ57L 62 est TD T92 RA73 102 TD T93 RZ26 87 TD T94 RZ36 88 est TD T95 RZ74 78 est TD T99 RZ27 88 est You might want to change processor specific thresholds in the Performa
99. a panel to connect automatically When you enable automatic connection for a panel it will to connect when you open the panel You can have several panels set for automatic connection To enable a panel to connect automatically 1 Select a panel from the Instrument Panel Directory in the Panel Manager by clicking on the panel name 2 Choose Enable from the Auto Connect menu The Instrument Panel Directory is updated The label auto connect is appended to the panel name To disable a panel from connecting automatically 1 Select a panel from the Instrument Panel Directory in the Panel Manager by clicking on the panel name 2 Choose Disable from the Auto Connect menu The Instrument Panel Directory is updated The label auto connect appended to the panel name is removed 344 Performance Manager Administrator Guide Terminate the Session Terminate the Session To end a DECwindows session 1 Pull down the File menu in the Panel Manager window and click the Exit menu item If you have modified panels Performance Manager displays the following message Panels modified but not saved Do you really want to exit 2 Click Yes and your changes are not saved OR Click No and the exit process is aborted 3 To save a panel use the Save or Save As menu entry in the panel See the section Set the Panel Options see page 366 section for more information about saving a panel How You Edit the Panel In
100. a working set required for the execution of a process for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MMG CTLFLAGPSPAR F CTLFLAGS The value of the SYSGEN parameter MMG CTLFLAGS which sets the target system memory management control settings for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MPW HILIMITPAR F MPW HILIMIT The value of the SYSGEN parameter MPW HILIMIT which sets un upper limit for the modified page list for the local node for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MPW LOLIMITPAR F MPW LOLIMIT The value of the SYSGEN parameter MPW LOLIMIT which sets the lower limit for the modified page list for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 510 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN MPW THRESHPAR F MPW THRESH The value of the SYSGEN parameter MPW THRESH which sets the lower bound of pages that must exist on the modifie
101. and the start transaction for the next terminal read or an image termination Channel TAPIO TAPTP TERM INPUT TERM THRUPUT VASIZE WSSIZE Rule CLUSTER OCCURRENCES CPU OCCURRENCES IO OCCURRENCES MEMORY OCCURRENCES OCCURRENCES POOL OCCURRENCES RESOURCE OCCURRENCES SCS BDT W BLKS R ADVISE PERFORMANCE GRAPH Description Number of process tape I O operations per second Number of process bytes per second transferred to and from tapes Number of process terminal read operations per second Number of process bytes per second transferred via terminal reads Number of pages in the virtual address space for a given process Number of working set pages X 1000 per process Description Rule Metrics available from history data only Number of rules prefixed with the letter L that fired per hour Does not include any rules in Domain Cluster Number of rules prefixed with the letter C that fired per hour Number of rules prefixed with the letter I that fired per hour Number of rules prefixed with the letter M that fired per hour Number of rules that fired per hour including user written rules Number of rules in the set R0020 R0025 R0030 R0035 R0040 R0045 R0050 R0060 R0070 R0080 that fired per hour Number of rules prefixed with an R but not in the above set that fired per hour Description Number of times per second that message had to wait
102. be experienced when longer queues form The format of this conclusion 10 spaces at the beginning of each line ensures consistent formatting with Performance Manager factory rules Brief Conclusion Text Element A brief conclusion is one line of text presented in the Brief Analysis Report when a rule fires A brief conclusion describes the problem detected by the rule The format for a brief conclusion is the following code BRIEF CONCLUSION text string To ensure consistent formatting with Performance Manager factory rules the text must be aligned with the left quotation mark Data Cell Types and Use The Performance Manager sub record types in the following table are stored in Performance Manager data records You can access the fields within the sub records as data cells within a rule construct See Appendix C for the names of the data cells You can also dump them using the ADVISE COLLECT REPORT DUMP DATACELLS command This command produces a dump report of all data cells in the LOCAL domain The following table shows the data sub record types and associated domains Sub record Type Associated Domain Communication COMMUNICATION Configuration CONFIGURATION CPU CPU 156 Performance Manager Administrator Guide Data Cell Types and Use Sub record Type Associated Domain Disk DISK Hot Files FILE Metrics COMMUNICATION CONFIGURATION CPU DISK FILE LOCAL PROCESS TAPE Parameters COMMUNICATION CONFIGURATION CPU DIS
103. being exported It must be a multiple of the interval of the data file and are rounded up to match such a value The default value is the interval of the data file being exported The value is expressed in seconds NODE NAME node name Identifies the node for which data is to be exported NODE NAME is required if your collection definition supports multiple nodes Only one node s data can be written out to an export file OUTPUT file spec Specifies the name of the export file If you do not specify OUTPUT or if you specify OUTPUT without a file specification a default filename of PSPASDUMP DAT is used 186 Performance Manager Administrator Guide ADVISE PERFORMANCE EXPORT PROCESS STATISTICS PRIMARY KEY option L SECONDARY KEY option This qualifier allows you to specify how process records can be summarized and sorted and to indicate what output fields to preserve in the output By default all details are preserved and all fields are supplied The default settings are PROCESS STATISTICS PRIMARY KEY PID SECONDARY KEY IMAGENAME If you use CLASSIFY BY to attach workload information to the process records you affect the default PRIMARY KEY which then becomes WORKLOAD NAME To obtain all detail fields such as PID including the workload name you must specify both the CLASSIFY BY qualifier and the PROCESS STATISTICS PRIMARY KEY PID SECONDARY KEY IMAGENAME settings Primary key options are
104. close the window pull down the File menu and release on the Close menu item By default the Full Analysis Report lists conclusions evidence and rule conditions that satisfied rule firings To remove either evidence or conditions or both from your report pull down the View menu and release on the appropriate item To proceed through the list of nodes click the appropriate Node arrow button or hold MB1 down on the node button to display an option menu with the selected nodes Releasing on the desired node name updates the window with that node s report Chapter 8 Use the DECwindows Motif Interface 279 How You Display Analyzed Data If more than eight nodes were selected for analysis the node option menu contains two more buttons which cause the node names to be shifted up or down when the cursor is within their option entry To stop the shifting move the cursor out of the more button Releasing on a node name causes the Analysis Report window to be updated with that node s report Releasing on a more button causes the report window to be updated with data of the node name adjacent to the more button To proceed through the list of conclusions click the appropriate rule ID arrow button Performance Evaluation Report Release on the Performance Evaluation menu item from the Display menu to open the Performance Manager Performance Evaluation Window as shown in the following screen File Customize SCS Statistics
105. configuration COMPAT EXEC FILE SYS IDLE INTERRUPT KERNEL MP SYNCH SUPER USER MODE Sampled process counts by process state COLPG MWAIT CEF PFW CUR LEF LEFO HIB HIBO SUSP SUSPO FPG COM COMO INPROCACT INPROCINACT OUTPROCACT OUTPROCINACT TOTAL PROCESSES Sampled process counts by type BATCH PROCESSES INTERACTIVE PROCESSES NETWORK PROCESSES DETACHED PROCESSES Derived CPU time metrics CPU Utilization Compute Queue CPU BATCH CPU INTERACTIVE BATCH COMQ INT COMQ 128 Performance Manager Administrator Guide Generate Custom Graphs CPU NETWORK CPU OTHER NETWORK COMQ DETACHED COMQ CPU DETACHED CPU TOTAL CPU MP INT Sampled paging swapping memory metrics Paging Rates MemoryPages Swaprate DZROFAULTS FAULTS FREEFAULTS FREECNT INSWAP GVALID MFYFAULTS PREADIO MPYCNT PREADS PWRITES PWRITIO SYSFAULTS WRTINPROG IMAGE ACTIVATIONS Derived paging swapping memory metrics Memory Utilization as a percentage Pagefile Utilization FREELIM FREELIST MODIFIED PAGEFILE UTILIZATION MEM TOTAL SYSTEMWS USERWS VMSALLOC Sampled system IO rates BUFIO DIRIO ERASE QIO FILE CACHE HIT FILE CACHE MISS FILE OPEN LOGNAM MBREADS MBWRITES SPLITIO WINDOW TURN Derived system IO rates Disks Terminals and Printers DISK PAGING DISK SWAPPING LAT TERMIO NV TERMIO RT TERMIO DISK USER TT TERMIO TX TERMIO TW TERMIO WT TERMIO OTHERBUFIO Sampled DECnet metrics ARRLOCPK ARRTRAPK DEPLOCPK RCVBUFFL TRCNGLOS Sampled distributed locking metrics
106. copies of the Documentation have been returned to CA or destroyed EXCEPT AS OTHERWISE STATED IN THE APPLICABLE LICENSE AGREEMENT TO THE EXTENT PERMITTED BY APPLICABLE LAW CA PROVIDES THIS DOCUMENTATION AS IS WITHOUT WARRANTY OF ANY KIND INCLUDING WITHOUT LIMITATION ANY IMPLIED WARRANTIES OF MERCHANTABILITY FITNESS FOR A PARTICULAR PURPOSE OR NONINFRINGEMENT IN NO EVENT WILL CA BE LIABLE TO THE END USER OR ANY THIRD PARTY FOR ANY LOSS OR DAMAGE DIRECT OR INDIRECT FROM THE USE OF THIS DOCUMENTATION INCLUDING WITHOUT LIMITATION LOST PROFITS BUSINESS INTERRUPTION GOODWILL OR LOST DATA EVEN IF CA IS EXPRESSLY ADVISED OF SUCH LOSS OR DAMAGE The use of any product referenced in the Documentation is governed by the end user s applicable license agreement The manufacturer of this Documentation is CA Provided with Restricted Rights Use duplication or disclosure by the United States Government is subject to the restrictions set forth in FAR Sections 12 212 52 227 14 and 52 227 19 c 1 2 and DFARS Section 252 227 7014 b 3 as applicable or their successors All trademarks trade names service marks and logos referenced herein belong to their respective companies Copyright 2008 CA All rights reserved Contact Technical Support For online technical assistance and a complete list of locations primary service hours and telephone numbers contact Technical Support at http ca com support Contents Con
107. current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 467 Performance Manager Data Cells MAX NONPAGEDPOOLBYTES INUSE Derived This contains a value representing the maximum number of non paged pool bytes in use on the local node of all of the intervals Data Type NUMERIC Domains SUMMARY MEMORY PAGES NOT ALLOC TO VMSMET F USERPAGES This contains the number of user memory pages for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP MODIFIEDLIST FAULT RATEMET F MFYFLTS Average number of pagefaults per second from the modified page list for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP MP_SYNCHMET_F_MP_SYNCH Average percentage of CPU time spent in MP synchronization for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP MULTI_IOMET_F_SPMIOBUSY Percentage of time that two or more of the disk devices were busy for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NETWORK_COUNTMET_F_NETWORK This contains a value representing the average number of network processes on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 468 Performance Manager Administrator Guide
108. data with DATATRIEVE interactively The DATATRIEVE record definition is contained in PSPASEXAMPLE PSPASDUMP BINARY DTR and PSPASEXAMPLE PSPASDUMP ASCII DTR 190 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH ADVISE PERFORMANCE GRAPH Format Description Qualifiers Use the ADVISE PERFORMANCE GRAPH command to chronologically graph any group of metrics stored in the Performance Manager database ADVISE PERFORMANCE GRAPH The Performance Manager can produce a multitude of predefined graphs You can also define your own custom graphs if the predefined graphs do not meet your specific needs You can generate multiple graphs with a single command The Performance Manager produces at least one graph and can produce many graphs depending on how you specify the TYPE and NODE qualifiers When producing many graphs you are likely to benefit by using the command mode interface invoked by the ADVISE PERFORMANCE command Command mode allows many graphs to be built in memory and selectively viewed or written to a file You can also direct the output to either the SYSSOUTPUT device or an output device as specified by the OUTPUT qualifier The Performance Manager generates the graphs using ReGIS format if the SYS OUTPUT device has the ReGIS capability or if you specify FORMAT ReGIS Otherwise the Performance Manager generates a less resolute graph using standard ASCII characters ANERAGE DAILY WEEKLY MONTHLY
109. displays The lower portion of a display common to all three displays is described in the following section 386 Performance Manager Administrator Guide Display RESOURCE Information Lower Common Resource Display The lower or common portion of each resource display contains bar graphs for nodes in the cluster An example is the lower portion of the next illustration The scale at left and right is graduated from O to 100 and is interpreted either as a percentage or an absolute value depending on the particular metric There is a bar graph for each node in the current group set for display the node name being displayed at the bottom of each graph The bar graph for a node contains three separate columns metrics as follows Memory Utilization M This column is headed by the letter M and is a percentage value as indicated by the percent sign 96 at the bottom of the column The value is that of Total MEMutl expressed as a percentage Total MEMutl is given by CTotal Memory Free Pages Total Memory I O Rate I This column is headed by the letter I and is an absolute value rate as indicated by the letter R at the bottom of the column The value is the Direct I O rate number of direct I Os per second for the node If the rate exceeds 100 direct I Os per second the column is filled with asterisks CPU Utilization C This column is headed by the letter C and is a percentage value as indicated by the percent s
110. each of the specified users or metrics separately on the plot To graph process metrics by user in DCL mode issue the command ADVISE PERFORMANCE GRAPH _ TYPE CUSTOM USER METRICS CPUTIME _ SELECTION DIEGER HAFFMAN ORIPAS _ SANTELK FRED CARL SARBID _ TITLE CPU time used by the team 130 Performance Manager Administrator Guide Generate Custom Graphs The Prompting for Process Custom Graph example shows Performance Manager prompts and user input that generate the same graph using command mode and specifying the following command PSPA GRAPH TYPE PROMPT FORMAT POSTSCRIPT CHARACTERISTIC PATTERN _PSPA gt OUT CHACUSTOM USER PROMPT PS The graph itself appears in the figure Custom Graph for Process Usage The following statements are keyed to the Prompting for Process Custom Graph example 1 2 3 4 5 2 was entered to select process metrics selected by user name The user names were entered to select the users CR was entered to request the list of process metrics CPUTIME was selected from the list and entered The Custom graph title was entered Chapter 4 Generate Historical Graphs 131 Generate Custom Graphs Please select either 1 a predefined graph or 2 a custom graph Choice 1 2 0 Composite 1 YQUEM 2 GALLO Please select a node number 0 1 For the CUSTOM Graph select one of the following System Metrics Process metrics selected by user name Process me
111. file must be a series of user fields separated by white space or commas User names can contain wildcard characters User names longer than 12 characters are truncated to 12 characters to ensure a match because the Performance Agent compares up to 12 user name characters only Account names can be up to 8 characters in length and process names up to 15 If you wish to preserve lower case characters enclose your entries in double quotes User criteria can be specified in terms of UICs account names process names or user names A UIC group can be indicated by using an asterisk for the user number 200 Control Buttons The Clear button removes all dialog box entries The Create button adds the workload name to the list and clears the entries Chapter 8 Use the DECwindows Motif Interface 311 How You Customize Delete a Workload Definition To delete a workload definition 1 Click a workload definition name The dialog box is updated to show the current definition field settings 2 Click the Delete control button to remove the workload definition The Performance Manager removes the workload definition name from the workload list and clears the definition fields If the definition is not deleted a message box is displayed explaining why the request cannot be executed A failure can occur when a workload family has been defined in terms of this workload A list of those families displays 3 Modify the workload family
112. files having different directories or devices you must define the logical names PSPASGRAPH FILE DIRECTORY and PSPASGRAPH FILE DEVICE Or alternatively use the FILTER qualifier to select hotfile records for specific processes and or disk devices If the graph legend is undesirable due to its length you have several options to shrink the font used to print the legend DECwindows has a Resources file that can change the font PostScript has a logical name The Pie Chart can be presented in tabular format which gives a wider legend text If the file is deleted before the prod fam Performance Manager detects its specification its name is not available The FID is provided in parentheses instead of its name All non virtual QIO activity to the disk is reported under the file specification Non Virtual QIO Chapter 4 Generate Historical Graphs 133 Chapter 5 Customize the Knowledge Base This chapter provides information about the knowledge base rules and customizing both This section contains the following topics The Knowledge Base see page 135 Investigate Rule Firing see page 136 Components of Rules see page 137 Data Cell Types and Use see page 156 Implement Changes see page 163 Build an Auxiliary Knowledge Base see page 172 Use an Auxiliary Knowledge Base for Reporting and Archiving see page 173 The Knowledge Base The Performance Manager makes assertions and inferences about syst
113. filespec Specifies that a file containing a series of date ranges is to be used in place of the BEGINNING and ENDING qualifiers Each line in the dates file should look as follows dd mmm yyyy hh mm ss cc dd mmm yyyy hh mm ss cc The time can be either omitted entirely or truncated Any truncated parts of the time default to O The periods of time represented by each line in the file need not be contiguous but they must be in ascending order DATES is incompatible with the BEGINNING and ENDING qualifiers ENDING time Specifies the ending date and time of the data to be exported If you do not specify BEGINNING the default ending time is the current time If you specify BEGINNING the default ending time is midnight 23 59 for the same day FILTER keyword The FILTER qualifier allows you select a subset of the daily or history data for exporting Process data and disk data can be filtered Process data can be filtered by using any of the filter keywords USERNAMES IMAGENAMES PROCESSNAMES ACCOUNTNAMES UICS PIDS or WORKLOADNAMES If a process record s identification information matches any of the identification specifications that are specified then that record is selected Likewise disk data can be filtered by using either of the filter keywords VOLUMENAMES and DEVICENAMES If a device record s identification information matches any of the volume names or device names that are specified then that record is selected
114. follows A decimal value is a rational number A literal symbol is a symbol previously defined by the literal construct A tally data cell is a data cell see the Data Cell Types and Use section Parentheses can be used to denote precedence A numeric data cell is a data cell see the Data Cell Types and Use section An index specifier data cell is a data cell see the Data Cell Types and Use section A Boolean data cell is a data cell see the Data Cell Types and Use section A scan routine data cell is a data cell see the Data Cell Types and Use section Chapter 5 Customize the Knowledge Base 153 Components of Rules Occurrence Element Evidence Element Present the Evidence m A string item is either a string data cell see the Data Cell Types and Use section or a string literal where the literal string is enclosed in quotation marks m A string operator is either the EQS or NES operator m Anumeric binary operator is one of the operators in the previous table When the Performance Manager evaluates a rule and all of the conditions of the rule are true there is a rule occurrence The number of times the Performance Manager evaluates a rule depends upon its domain the number of interval records and the number of nodes The format for an occurrence element is as follows OCCURRENCES literal symbol decimal value A literal symbol or decimal value specifies the number of times a rule must occur f
115. for buffers Block request rate Chapter 6 Performance Manager Commands 209 ADVISE PERFORMANCE GRAPH Channel BLKS S CR W DGS D DGS R DGS S KB MAP KB RC KB S MGS R MGS S System ARRLOCPK ARRTRAPK BATCH COMQ BATCH PROCESSES BUFIO CEF COLPG COM COMO COMPAT CPU BATCH CPU DETACHED CPU INTERACTIVE 210 Performance Manager Administrator Guide Description Block send rate Number of times per second that messages had to wait due to insufficient credits Datagrams discarded rate Datagram receive rate Datagram send rate Kbytes transferred rate Kbytes received rate Kbytes sent rate Message receive rate Message send rate Description Arriving local packets per second Transit packets per second Number of computable batch processes Number of Batch processes Buffered I O per second Average number of processes in common event flag wait state Average number of processes in collided page wait state Average number of processes in computable state Average number of processes in computable outswapped state Percent CPU time spent in compatibility mode Percent CPU time used by batch jobs Percent CPU time used by detached jobs Percent CPU time used by interactive jobs Channel CPU NETWORK CPU OTHER CPU TOTAL CPU VUP RATING SWPBUSY IOBUSY ANYIOBUSY PAGEWAIT SWAPWAIT MMGWAIT SYSIDLE CPUONLY IOONLY CPUIO CUR DEADLOCK FIND DEADLOCK SEARCH DEPLOCPK DETA
116. interval with the INTERVAL qualifier BYCLUSTER Each Tabular report section is presented in a cluster wide format The configuration section is not available in the cluster wide format BYNODE Each Tabular report section is presented in a cluster wide format with the by node detail included The configuration section is not available in the cluster wide format Chapter 6 Performance Manager Commands 235 ADVISE PERFORMANCE REPORT Examples If both BYNODE and BYCLUSTER are omitted the Tabular report sections are presented on a node by node basis and not on a cluster wide basis By default if none of the above options are specified FINAL is assumed Using BYCLUSTER or BYNODE presents a different output format than FINAL or INTERVAL For example the following commands produce different output formats ADVISE PERF REPORT TABULAR SECTION ALL ADVISE PERF REPORT TABULAR BYCLUSTER SECTION ALL ADVISE PERFORMANCE REPORT ANALYSIS PERFORMANCE EVALUATION _ OUTPUT SAMPLE This command produces an Analysis Report and a Performance Evaluation Report for the current day using a beginning time of midnight 00 00 and the current time of day as the ending time The reports contain information for each of the nodes listed in the Performance Manager schedule file The OUTPUT qualifier directs the output to a file called SAMPLE RPT The RPT extension is the default ADVISE PERFORMANCE REPORT ANALYSIS NODE NAMES DEMAND _ NOEXP
117. is FREEBLKS TYPE TOP FREEBLK VOLUMES Top Freeblks Disk Volume Plots the top five plus Other disk volumes in terms of number of free disk pages The metric graphed is FREEBLKS 222 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE TOP HARDFAULTING IMAGES Top Hard Faulting Images Plots the top five plus Other Images creators of hard page faults by image name The metric graphed is HARDFAULTS TYPE TOP HARDFAULTING USERS Top Hard Faulting Users Plots the top five plus Other Users creators of hard page faults by user name The metric graphed is HARDFAULTS TYPE TOP HARDFAULTING WORKLOADS Top Hard Faulting Workloads Plots the top five plus Other Users creators of hard page faults by workload name The metric graphed is HARDFAULTS TYPE TOP HSC DISK IO Top HSC Disk IO Plots the top five plus Other HSCs in terms of disk I O operations per second The metric graphed is HSC DISK IO TYPE TOP HSC DISK THRUPUT Top HSC Disk Thruput Plots the top five plus Other HSCs in terms of disk throughput in Kilobytes per second The metric graphed is HSC DISK THRUPUT TYPE TOP HSC IO Top HSC IO Plots the top five plus Other HSCs in terms of I O operations per second The metric graphed is HSC IO TYPE TOP HSC TAPE IO Top HSC Tape IO Plots the top five plus Other HSCs in terms of tape I O operations per second The metric graphed is HSC TAPE IO TY
118. is not Data Type NUMERIC Domains CFG SCS NODE ON RFCFG V STATUS RF This contains a Boolean of either the values 1 if the remote node for the current configuration subrecord is accessed over an RF controller or a O if it is not Data Type NUMERIC Domains CFG Appendix C Performance Manager Data Cells 497 Performance Manager Data Cells SCS PATHNAMECFG A PATH This contains the device name string for the path over which the local node has SCS communications with the remote node indicated by SCS NODENAME for the current configuration subrecord e g PAAO or PEAO Data Type STRING Domains CFG SEND CREDIT QUEUE RATESCS F QCR CNT This contains the value representing the number of times per second that SCS messages had to be queued on the local node that were destined for the remote node indicated by SCS NODENAME for the current configuration record and interval Data Type NUMERIC Domains CFG SEND CREDIT QUEUE TALLY Derived This contains the sum of the values representing the number of times per second that an SCS message had to be queued on the local node that was destined for the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG SEQUENCED MESSAGES RECD TALLY Derived This contains the sum of the values representing the number of messages
119. local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP Appendix C Performance Manager Data Cells 515 Performance Manager Data Cells SYSGEN SMP CPUPSPAR F SMP CPUS The value of the SYSGEN parameter SMP CPUS which sets which secondary processors if available are to be booted into the multiprocessing system at boot time for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN SPTREQPAR F SPTREQ The value of the SYSGEN parameter SPTREQ which sets the number of system page table SPT entries required for mapping the OpenVMS Executive image RMS image SYSMSG EXE file multiport memory structures and other OpenVMS components for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN SRPCOUNTPAR F SRPCOUNT The value of the SYSGEN parameter SRPCOUNT which sets the number of pre allocated small request packets for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN SRPCO
120. minimum average and maximum number of global pages in use by the process at the reporting interval Priv MIN AVE MAX The minimum average and maximum number of process private pages in use by the process at the reporting interval 92 Performance Manager Administrator Guide Tabular Report Sections WS Defit The working set default value for the process WS Quota The working set quota value for the process WS Extnt The working set extent value for the process Virt MIN AVE MAX The minimum average and maximum virtual page count for the process during the reporting interval Tabular Report YQUEM VAX 6000 440 PA Vx x Tuesday 26 JAN 2006 09 00 to 10 00 kakatakakatakatatakalatalakatakatatakakatakakok Node YQUEM Final Statistics katakakatakakatalakatalakatalalatakatatakatak EERE EEE TAREE AER Data Analyzed from 26 JAN 2006 09 00 00 00 to 26 JAN 2006 10 00 00 00 AA EER RAE MIN WS MAX User Name Account Globl i i Deflt i Virt virt 29400201 SYSTEM 29400206 SYSTEM start 29400209 SYSTEM start 2940020A SYSTEM start 2940027E MACNEIL 2940027F FRIES 294002C3 LUNDGREN 294002EC STEPHENS 294010E4 BHAT 294010E5 RAMAN 2940115B FRIES 29401286 DBIGELOW 29401EF9 login login 29401EFB OPERATOR CNB 29401EFC BHAT 341 29401EFE DQS SERVER CNB User Command ADVISE PERF REPORT TAB SECT EXTEND NODE YQUEM BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 10 00 00 00 OUT TAB
121. of these words begins a line of memory statistics For Cluster Total the value is left blank in case of percentages Node Average The average value of the statistic across all nodes in the cluster system Node Minimum The minimum value of the statistic across all nodes in the cluster system 94 Performance Manager Administrator Guide Tabular Report Sections Node Maximum The maximum value of the statistic across all nodes in the cluster system Cluster Total The total value of the statistic across all nodes in the cluster system By Node Format The following example shows final statistics by node Tabular Report CLUSTER PA Vx x Tuesday 26 JAN 2006 09 00 to 09 20 SR RK KK KK Final Statistics SESE SORA kk Data Analyzed from 26 JAN 2006 09 00 00 00 to 26 JAN 2006 09 20 00 00 FEO AR AAA KAA AA FEES EAA SK ARK Total Memory Proc Balset Page Hard Soft Gvalid InSwP MEMutl Queue Count Count Faults Faults Faults Faults Count avg sec LATOUR CLUSTER I 0 CPU System CPU CPU CPU IO I 0 Multi CPU IO Direct Buffrd Busy CPU Queue Idle Idle Only 1 0 Busy I 0s 55 LATOUR CLUSTER H Orig Out Lck Act Bound 55 LATOUR YQUEM User Command ADVISE PERF REPORT TABULAR BYNODE BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 09 20 00 00 OUT BYCLU TAB SECTION SUMMARY In the By Node format for summary statistics the leftmost column of the tabular report contains the name of each node
122. of thresholds To modify a threshold value m Click the threshold field and type the new value The x axis on the strip chart can be changed by specifying a new value for the Number of Time Units This value with the display interval determines the amount of time represented on the strip chart To apply your changes 1 Click the Apply button at the bottom of the dialog box 2 When you are satisfied with all your changes and want to exit the dialog box click OK to exit without saving any changes click Cancel Change the Colors and Patterns You can customize the colors used in any part of a PA Real time Display panel or instrument To change the high threshold color for the strip chart in the CPU panel 1 Select the instrument by pressing MB1 within the instrument s border The instrument will appear depressed to indicate that it has been selected 2 Press MB3 and hold to get the pop up menu 3 Choose Colors and Patterns by moving the cursor over this selection and releasing the MB3 button The following dialog box is displayed Chart Background Normal Range Low Threshold High Threshold This dialog box lets you set colors for the thresholds set in the Ranges and Thresholds dialog You can use these colors to alert you to potential performance problems Chapter 9 Use the DECwindows Motif Real time Display 339 Change the Colors and Patterns 4 Alter the background or the threshold colors by clicking o
123. one was detected Disk Avg I O Avg Avg I0sz Source Service IO IO Volume per Sec Queue Kb sec in pgs Node Node Busy Read Split Type Samples ARNOLD 2 DJA4 0 00 O 0 00 O 0 00 O 0 00 O 2 DUA15 0 03 0 00 0 03 0 00 SUPPLY DEMAND NODE1 2 0 2 0 2 DUA5 0 06 5 r 0 00 O 6 SUPPLY 0 04 3 DEMAND 0 01 1 NODE1 2 DUA10 0 14 16 0 00 O 0 SUPPLY 0 08 9 2 DEMAND 0 06 7 9 NODE1 2 DUA7 0 40 x 0 01 4 SUPPLY 0 35 x 3 DEMAND 0 04 0 NODE1 The following statements are keyed to the columns in the previous example 1 Volume name of the disk to which one or more nodes in the cluster directs activity 2 Average number of I O operations per second to the disk volume For each disk the total I O per second is reported followed by a breakdown of this activity from contributing nodes when applicable In this example there was an average of 2 02 I O operations per second to disk volume CLEM with node SUPPLY contributing an average of 0 14 I O operations per second node DEMAND contributing 1 46 and node NODE1 accounting for the remaining 0 42 I O operations per second 3 Average number of I O requests waiting for service to the disk The average queue size for each disk is followed by a breakdown of this value from contributing nodes when applicable Chapter 3 Evaluate Performance in Detail 57 Performance Evaluation Report 4 Average number of kilobytes per second t
124. or processes access most Note that the I O rate might be understated and that some other disk might be accessed heavily but not reported because the primary data collector is only capturing the top two devices that the process uses each two minute interval If an alternate collection is used these columns show no data since the data is not provided by an alternate collector m 2nd Top Disk Volume Name IO sec See previous paragraph m 3rd Top Dsk Volume Name IO sec See previous paragraph View Process Statistics with an Emphasis on Memory Metrics The following report example illustrates the process statistics section with a focus on process memory related metrics Some relevant UAF parameters are also displayed The example uses the default primary and secondary keys of IMAGENAME and USERNAME This format of the report is obtained by using the qualifier PROCESS STATISTICS FOCUS MEMORY RELATED To view the report with a different orientation provide your choice of sort keys For example the following syntax presents the statistics by user with a breakdown by image PROCESS STATISTICS FOCUS MEMORY RELATED PRIMARY KEY USERNAME SECONDARY KEY IMAG ENAME 48 Performance Manager Administrator Guide Performance Evaluation Report Performance YQUEM VAX 6000 440 Evaluation Tuesday 26 JAN 2006 09 00 to 10 00 Primary Secondary Key SftFlt HrdFlt Ave Ave Ave Obs Max Max Ave Ave WS WS Image User Sec Private Global VA Spac PFW VA Sp
125. pages average private pages maximum private pages working set default working set quota working set extent minimum virtual size average virtual size maximum virtual size Position 215 224 233 239 255 266 275 284 293 302 311 320 329 338 347 356 365 Length 7 12 Appendix E Output Format for ASCII CSV Data 557 Process Metric Statistics Data Record Example Data item 1 DUA3 image dev SYS5 SYSCOM image dir MON SYSEXE SET image name UTILITIES workload name Description image device name image name directory image file name the workload name provided with CLASSIFY qualifier Position 375 404 445 488 Length 25 37 39 18 558 Performance Manager Administrator Guide Appendix F How You Graph Seven or More CPUs Performance Manager can graph six CPUs per command If you need to create a graph that displays seven or more CPUs for a node you need to write CSV files for 6 CPUs at a time manually create a merged CSV file move the file to a Windows machine and then use Microsoft Excel to create a graph To graph seven or more CPU s complete the following tasks in order 1 Create a CSV file 2 Create More CSV Files as Necessary 3 Create a Single CSV File 4 Send the CSV file to a Windows Machine 5 Create the graph in Excel This section contains the following topics
126. per second transferred from all hot files indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME for the current hot file subrecords which were selected by the most recent FILE_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FILE_SCAN Derived Provides the count of hot file subrecords for which the specified rule condition is true The condition will be evaluated for each hot file subrecord Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FILE SPLIT IO RATEFIL F SPLITS This contains a value indicating the number of split I O operations per second issued to the file indicated by FILE DEVICE FILE DIRECTORY and FILE NAME for the current interval Data Type NUMERIC 450 Performance Manager Administrator Guide Performance Manager Data Cells Domains FIL FILE SPLIT IO TALLY Derived This contains the sum of the values indicating the number of Split I Os per second transferred to and from all hot files indicated by FILE DEVICE FILE DIRECTORY and FILE NAME for the current hot file subrecords which were selected by the most recent FILE SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FILE_SWAPPING_IO_RATEFIL_F_SWPOP This contains a value indicating the number of I O operations per second issued by the SWAPPER to the file indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME
127. problems it does not by default screen out firings that are insignificant Some circumstances on your system might fire a rule for a one time transient condition without any implications for the long term performance of your system The factory rules try to minimize this possibility by setting occurrence thresholds within rules but this might still happen Figure out whether you need to be concerned about a particular rule firing Check the evidence provided with each rule Does the problem appear to be persistent as indicated by many lines of evidence Check the time of occurrence does this happen at the same time as a regularly scheduled job Make sure you understand the meaning of each data item presented If you are unsure of a definition look it up in the Data Cell Types and Use section or use the online Help system which contains hotspots to all the data items in the Conditions and in the Evidence Many of the data items presented as evidence are used in the rule s conditions read the conditions and understand how each data item is being used and what values would cause it to fire You are probably very familiar with approximate or typical values for many parameters of your systems If a rule fires but all related evidence appears normal for your environment then you probably want to adjust that rule s thresholds to reflect more precisely the upper bounds of performance for your particular workloads and systems For more informa
128. process 398 Performance Manager Administrator Guide Evaluate Performance Using the Investigate Displays Large processes with swapping disabled Look at the Working Set and Process bars for inactive processes with large working sets If these processes have swapping disabled they cannot be swapped but retain memory at the expense of other processes Use the system dump analyzer SDA to see if a large inactive process has the PSWAPM prohibit swap mode bit set Inappropriate page cache sizes Look at the page fault rate Hard flt and Soft flt free memory Free pgs and swapping Working Set and Process bars If the overall fault rate is high and the faults are mostly soft faults the page cache might be too large This may also be accompanied by swapping and extensive free and modified page lists The page cache is encroaching on memory that could be made available for working sets If the overall faulting rate is low while the hard fault rate is high the page cache is ineffective that is the free page list and or modified page list is too small There is ample memory for working sets but the caching effectiveness is low The sizes of the page caches are controlled by the system parameters FREELIM FREEGOAL MPW LOLIMIT and MPW THRESH Isolate the Cause of an I O Limitation If an examination of the system overview reveals an I O limitation you can investigate the cause of the limitation in more detail using the I O di
129. quickly skipping around as dictated by your investigation without causing the input data to be reread and reanalyzed Also if you need to produce output files in different formats command mode is more efficient than using DCL commands 108 Performance Manager Administrator Guide Generate Predefined Graphs To invoke command mode m Enter the following command ADVISE PERFORMANCE PSPA gt In command mode you must first select the data you want to use and then you specify the method you want to use to view it The graphs can be viewed in ReGIS mode on a terminal and then written to a file in Postscript for example without having to reread and analyze the data Also you can switch between graphs and pie charts without reprocessing the data To select data enter the SELECT command and specify the processing options you want to use You can specify any or all of these processing options ANALYSIS PERFORMANCE and GRAPHS For example PSPA gt SELECT GRAPHS NODE SUPPLY This command selects data for node SUPPLY subsequent GRAPH commands use this data to generate graphs The ANALYSIS option provides the results of the factory rules and optionally user rules that may fire as a result of your performance data The PERFORMANCE_EVALUATION option allows you to view and output sections of the Performance Evaluation Report and Tabular Report which contain statistics about the system including process and disk activity and
130. rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DECNET TRANSIT PACKET RATEMET F ARRTRAPK Average rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DEMANDZERO_FAULT_RATEMET_F_DZROFLTS Average number of demand zero pagefaults per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DEPARTG_DECNET_PACKET_RATEMET_F_DEPLOCPK Average DECNET departing local packet rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DEVICE_NAMEDEV_A_DEVNAME This contains a string indicating the disk device of the disk for which the current disk subrecord pertains e g 2 DUA11 Data Type STRING Domains DSK 432 Performance Manager Administrator Guide Performance Manager Data Cells DIRECTORY DATA CACHE AR Derived Attempt rate per second to the directory data cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DIRECTORY DATA CACHE HR Derived Hit ratio to the directory data cache for the local node for the current interval record Calculated by dividing the number of directory data cache hits by the number of directory data cache attempts hits misses times
131. replaced by a model number All Integrity servers use 4096 for a model number For Alpha systems you can get the model number of a specific processor by using the GETSYI system service using the item code SYI HWMODEL An entry of zero is used for an unknown model A scale factor of 1 0 is used as a basis for the rules when applied to a VAX 11 780 system For example the threshold construct TD SOFT FAULT SCALING 40 0 4 places the value 0 4 in the data cell SOFT FAULT SCALING when a VAX model 40 is being analyzed A VAX model 40 is a MicroVAX 2000 Chapter 5 Customize the Knowledge Base 147 Components of Rules Rule Construct Elements A rule construct consists of up to seven elements Each construct begins with RULE and ends with ENDRULE The following sections describe each of the elements listed in the following table Rule Construct Elements and Descriptions Element Rule ID Domain Rule conditions Occurrence Evidence Conclusion text Brief conclusion text Presence Required Optional Required Optional Optional Optional Optional Description Identifies a rule definition Identifies the domain default is LOCAL There must be at least one condition Number of occurrences for the rule to fire default is one If omitted no evidence is presented If omitted no conclusion text is presented If omitted no conclusion text is presented Auxiliary Rules File Example As show
132. second have values of 20 23 22 and 27 thereby contributing to more than 50 percent of the total operations to volume BRANDY1 Miscellaneous Rule Analysis Report The following example shows a Miscellaneous Rule Analysis report Full Analysis SUPPLY VAX 11 780 Saturday 01 OCT 2006 00 00 to 23 59 CONCLUSION 4 R0010 The system fault rate for VMS is over 2 faults per second for the following time periods Performance can be improved for the whole system if the VMS fault rate can be reduced Increase the working set size for VMS SYSMWCNT to reduce the system fault rate Do this by adding an entry in MODPARAMS DAT similar to ADD SYSMWCNT 100 and running AUTOGEN The 100 is just an initial guideline Total number of samples giving this conclusion 12 Current setting of this system parameter 730 CONDITIONS 1 SYSTEM FAULT RATE GE 3 00 2 OCCURRENCES GE 4 EVIDENCE System fault Time of occurrence Ui Ui Ul Ui Ui Ui Ui Ui Ui Ui Ui Ui pi Chapter 2 Analyze Performance 27 Analysis Reports The following statements are keyed to the columns in the Miscellaneous Rule analysis report 1 Average number of times per second that a page fault hard or soft occurred for the OpenVMS system working set 2 Beginning time of the interval in which the condition occurred Analysis Summary Report The following example shows an Analysis Summary report Full Analysis SUPPLY VAX 11 780 Saturday 01 OCT 2006
133. shows the transfer counts and Kbytes mapped as I O that it initiated to satisfy requests for data made by a remote node This is the only time numbers for transfer counts and Kbytes transferred are reported The initiator of the transfer is not the node that issues the initial QIO but the node that issues the SCS directive for block mode transfer services to satisfy the I O request Sequenced messages are used by the Distributed Lock and Connection Managers also implicitly in disk I O to set up block mode transfers For sequenced messages SCS imposes its own flow control and delivery and order of messages is guaranteed Data G Sent Rate number per second at which datagrams are sent to the named node by the local node doing the data collection Data G Recvd Rate number per second at which datagrams are received from the named node by the local node doing the data collection Data G Discd Rate number per second at which datagrams are discarded by the CI port driver because a receive buffer is not available This is the rate at which datagrams are sent to but never received by the named node from the local node doing the data collection Msgs Sent Rate number per second at which sequenced messages are sent to the named node by the local node doing the data collection Msgs Recvd Rate number per second at which sequenced messages are received from the named node by the local node doing the data collection Snd Cr
134. the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN SWPRATEPAR F SWPRATE The value of the SYSGEN parameter SWPRATE which sets the swapping rate and serves to limit the consumption of disk bandwidth by swapping for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP Appendix C Performance Manager Data Cells 517 Performance Manager Data Cells SYSGEN SYSMWCNTPAR F SYSMWCNT The value of the SYSGEN parameter SYSMWCNT which is the system working set count which establishes the number of pages for the working set containing the currently resident pages of pageable system space for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN VBSS ENABLEPAR F VBSSENA The value of the SYSGEN parameter VBSS ENABLE which determines whether Virtual Balance Slots available with OpenVMS V6 0 and above are enabled for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN WSDECPAR F WSDEC The value of the SYSGEN parameter WSDEC that specifies t
135. the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWSCSPRO V RSN SCS This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWSCS mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWSWPPRO V RSN SWPFILE This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWSWP mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN SUSPPRO V SSS SUSP This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the suspended scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Appendix C Performance Manager Data Cells 493 Performance Manager Data Cells Domains PRO PROCESS WAS IN SUSPOPRO V SSS SUSPO This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the outswapped suspended scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCE
136. the following screen File Edit Display Display Top Display Custom TOP CP For more information on predefined graphs see the chapter Evaluate Performance in Detail see page 33 294 Performance Manager Administrator Guide How You Display Analyzed Data Display Custom Graphs Pull down the Display Custom menu to display custom graphs Choose the custom graph category that you want to select Performance Manager interface displays a dialog box as shown in the following screen Directions Select multiple metrics for one item or one metric and multiple items Metries Options Selections ANYIOEUSY ARRLOCPK ARRTRAPK BATCH COMQ BATCH PROCESSES BUFO mme You can take the following actions in this screen m To move your selected option into the selections box click a metric option Up to six system metrics can be selected m To apply your changes and draw the graph click the Apply button without closing the dialog box m To apply your changes and close the dialog box click the OK button m To clear all selected items click the Reset button m To close the dialog box without applying any changes click the Cancel button For most custom categories additional items appear in the lower option box You can display up to six metrics when using system metrics For other types of metrics you can display up to six metrics when one item is chosen and up to six items when one metric is c
137. the following command ADVISE PERFORMANCE COMPILE MYRULES VPR In this example MYRULES VPR is the name of your source rules file The Performance Manager compiles the rules and then names the resultant knowledge base MYRULES KB Note If you compile an auxiliary rules file it is valid only for the version of the Performance Manager software on which it was compiled and later versions Use caution to not place it in a cluster common area This is due to concern that the compiled rules file might be used on older software versions where the new keywords and data cells are not defined The filename PSPASKB KB is reserved by CA and should not be used to name an auxiliary rules file Use an Auxiliary Knowledge Base for Reporting and Archiving The Performance Manager generates Analysis Reports based on the Performance Manager factory knowledge base combined with the auxiliary knowledge base when you specify the name of your auxiliary rules file with the following command ADVISE PERFORMANCE REPORT ANALYSIS RULES MYRULES KB The Performance Manager archives analysis results for rules in all domains except CLUSTER and SUMMARY When you issue the following command rule occurrences based on the Performance Manager factory knowledge base combined with the auxiliary knowledge base are archived ADVISE ARCHIVE RULES MYRULES KB If you want to use the auxiliary knowledge base regularly you can use the Performance Manager Paramet
138. the metric CPU Utilization instrument in the Performance Manager System Overview panel launches the CPU panel on node YQUEM The node YQUEM and the current interval are passed to the panel CPU CPU on YQUEM CPU Only CPU amp VO Baw VO Only CPU amp VO kie CPU Utilization Top Users EGELMAN ORREY 3 256 papiro frm Chapter 10 Customize the DECwindows Motif Real time Display 365 How You Set the Panel Options Double clicking on the Top Users metric name SAPIRO launches the user CPU panel passing it the current interval the node YQUEM and the user name SAPIRO as shown in the following screen eS ugs PUE NI IER BT AR User SAPIRO Ee e e ele ea NA Mens File Commands Edit Options CPU Utilization CPU Utilization 2 40 Direct 10 Bate CT mom To am OO Herd Fault Rate L 1I Soft Fault Rate Ce Buf I O Rate L 7 User Processes User Images dehin WEB OA EBUGSHR prwors jJ 5 VOTESSMAIN Each of the metrics in the panel are filtered by the user name SAPIRO For example the User Processes instrument is set up as follows The instrument s main metric is Process CPU Utilization sorted in ascending order The metric is filtered by instance with a filter metric of user name The filter compare value is left blank so the value is passed from the panel The instrument s second metric is process name using an alias of User Processes How You Set the Panel Op
139. the percentage of CPU Utilization for the data line of each process is scaled according to the processor speeds of the nodes in the cluster The speed ratings can be changed using an auxiliary knowledge base If you specify CLUSTER or BY NODE a by node breakdown of the process data is provided following the line representing the cluster wide data SECTION item Specifies which sections of the Tabular Report should be displayed By default all are displayed For node analysis the available sections are as follows m CONFIGURATION m SUMMARY STATISTICS m DISK STATISTICS m PROCESS STATISTICS m EXTENDED PROCESS STATISTICS 254 Performance Manager Administrator Guide SAVE SAVE Format SPAWN Format Parameter For PROCESS STATISTICS and EXTENDED PROCESS STATISTICS the process data must be selected with the PROCESS ALL keyword For example PSPA gt SELECT PERFORMANCE PROCESS ALL For cluster analysis the available sections are as follows m SUMMARY STATISTICS m DISK STATISTICS The SAVE command allows you to save a selection of graph data to a disk file in a binary format All graph data points are saved and they can be reloaded using the LOAD command The SAVE operation does not affect the current selection SAVE file spec The SPAWN command creates a subprocess of the current process Portions of the current process context are copied to the subprocess SPAWN command string commandc string
140. the system wide direct I O rate including all disks If there is a high direct I O rate for your system the disks might be a bottleneck A value for direct I O rate that exceeds this number indicates a high direct I O rate Data in the disk bars shows how much of the disk s time is spent doing I O on behalf of the user A high rate for a given disk may indicate that the disk is a bottleneck Buffered I O high A high value of Buf I O the buffered I O rate may indicate an I O limitation If the value of Buf I O exceeds this number a high buffered I O rate is indicated Chapter 11 Use the Character Cell Real time Display 395 Evaluate Performance Using the Investigate Displays Investigate a CPU Limitation The chief indicators of a CPU limitation are as follows Processes in the CPU queue Look at the value of CPU queue and at processes in the COM CUR states in the Priority bars CPU que gives the number of processes waiting for the CPU COM state There is a real queue if this value is greater than 1 The process displaced by Performance Manager the NULL process is not counted The existence of a real CPU queue indicates a CPU limitation A significant number of processes in the COM CUR states COM computable waiting for the CPU CUR the current process also indicates a CPU queue and consequently a CPU limitation idle time Look at the Idle Idl bar in the center of the display If there is no CPU idle time t
141. the value of Dzero flts A large number of demand zero faults indicates an excessive number of image activations Activating an image in a process involves considerable overhead If Dzero faults is a large percentage of total faults Hard flt Soft flt image activations might be excessive Paging induced by image activations is unlikely to respond to system parameter changes Application design changes are needed Balance set too small Look at Proc cnt number of processes on system Balset number of processes in balance set Free pgs number of pages of free memory and swapped processes If the balance set count is too small processes are swapped even if there is still free memory If Balset is significantly less than Proc cnt and Free pgs is adequate then the balance set count is too low Set the system parameter BALSETCNT to a value two less than the system parameter MAXPROCESSCNT A few active processes consuming memory Look at the Process bars in particular for active processes with large working sets For example a low priority compute bound process is less likely to be swapped than one that performs terminal I O They may cause other processes to swap Decreasing DORMANTWAIT may help if the large processes are above their working set quotas You can also suspend the large process with SET PROCESS SUSPEND and allow the swapper to trim it back to SWPOUTPGCNT The underlying problem might be that WSQUOTA is too large for the
142. then click the Delete button The name is removed from the list box To remove multiple names click the names and then click the Delete button Chapter 8 Use the DECwindows Motif Interface 303 How You Customize To create a list of devices push down and hold on the uppermost box in this window which displays All devices or one of the other options When the menu appears pull down to the menu item you want and release Use option menu to select coverage mu or wo cancer ue Click the text entry field to activate the cursor To add a device name enter a name and press Return or click the Enter button The name will appear in the list box and be cleared from the text entry field To delete a name click the name in the name list box then click the Delete button The name is removed from the list box To remove multiple names click the names and then click the Delete button 304 Performance Manager Administrator Guide How You Customize The following sample Collection Coverage List box shows a list of processes for which no data is to be collected Use option menu to select coverage Exclude processes CI m The type of list being managed cannot be converted For example an include processes list cannot become an exclude processes list All list entries must be deleted before the coverage list type can be changed Customize the PSDCSDATABASE Definition To specify or redefine a PSDC DATABAS
143. time spent in Supervisor mode for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 501 Performance Manager Data Cells SWAPPER TRIMMING TOO SEVERE Derived This contains a Boolean value representing severe swapper trimming on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SWAP_BUSYMET_F_SPMSWPBUSY Percentage of time that the Swapper was busy for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SWAP_WAITMET_F_SPMSWAPWAIT Percentage of time that the CPU was idle and at least one disk device had swapping I O in progress for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SYSGEN_ACP_DINDXCACHEPAR_F_ACP_DINDXCACHE The value of the SYSGEN parameter ACP_DINDXCACHE which controls the size blocks of the directory index cache and the number of buffers used on a cache wide basis for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN_ACP_DIRCACHEPAR_F_ACP_DIRCACHE The value of the SYSGEN parameter ACP_DIRCACHE which sets the number of pages blocks for caching directory blocks f
144. to remove the reference to this workload definition Modify Workload Definitions To modify a workload definition 1 Click a workload definition name The dialog box will display the current definition values 2 Modify settings as you wish and click Modify at the bottom of the box 312 Performance Manager Administrator Guide How You Customize Workload Family Definitions To define or modify workload families 1 Pull down the Parameters submenu and choose the Workload Families menu item Performance Manager displays the Workload Family Definitions dialog box as shown in the following screen EACH ACCOUNT EACH PID EACH UIC GROUP EACH USER Workloads Excluded Included ABC IMAGES ALL USERS COMPILES gt COMPILES_IMAGE a COMPILES TERM DCL m I The menu bar contains Control and Help menus A list of workload family names appears at the top of the dialog box To close the dialog box pull down the Control menu drag the pointer to the Exit menu item and release MB1 The Workload Family Definitions dialog box lets you do the following actions Create a workload family Delete a workload family Modify a workload family Chapter 8 Use the DECwindows Motif Interface 313 How You Customize Create a Workload Family To create a workload family m Enter any of the following parameters Family Name Enter the name of the workload family The family name is limited to 20 chara
145. transfers that were initiated by the local node Used for block transfers primarily HSC MSCP and Connection Manager transfers This field is zero for all other nodes in the list A process on a remote node does a QIO write to a disk that is hosted locally The MSCP server on the local node has to do an SCS Request Data Directive that initiates a block mode transfer to read from the remote node so that it can write that data to the local disk K Byte Mapd The amount of buffer space in kilobytes mapped to receive data from or send data to the named node by the local node doing the data collection Used for block transfers primarily HSC MSCP and Connection Manager transfers Buf Dsc Queued Metric related to block transfers The number of times that a local node Sysap attempted to map a buffer and there were no free buffer descriptor table BDT entries available The number of BDTs is controlled by the system parameter SCSBUFCNT Chapter 3 Evaluate Performance in Detail 87 Tabular Report Sections Disk and Server Statistics Section The following example shows the Disk Statistics Section Tabular Report YQUEM VAX 6000 440 PA Vx x Tuesday 26 JAN 1997 09 00 to 10 00 FR A A A 2 2 A RE E E E E OK k A kkk kK Node YQUEM Final Statistics AR 2 2 2 A K K KK KKK kkk K KKK die ebeeee eoe e Data Analyzed from 26 JAN 1997 09 00 00 00 to 26 JAN 1997 10 00 00 00 eeeekebebeboeeeooekr Node YQUEM Rate Read Remote i Queue Space
146. used in a rule expansion 138 Performance Manager Administrator Guide COMMENT Components of Rules Constructs allow you to include notations in the rules file m An exclamation point occurring as the first character in the line denotes a comment construct The Performance Manager ignores any text on the line after the exclamation point THRESHOLD Constructs allow you to modify internal thresholds that affect the calculation of derived data cells m The format of a threshold construct is as follows THRESHOLD threshold definition threshold definition ENDTHRESHOLD m A threshold definition is defined as threshold name decimal value m A threshold name must match one of the Performance Manager s predefined thresholds These thresholds are not used directly in the rule expression however they influence the resultant value of some data cells that can be used in rule expressions Some thresholds act as occurrence limits for a few factory rules m The following table lists the Performance Manager s predefined thresholds and the default value of each threshold and describes how each threshold affects the data cells or evidence m All thresholds have a prefix of TD These thresholds cannot be redefined but their values can be modified The following tables show the Performance Manager thresholds Name Default Value TD MIN DSKSPC PCT 05 TD DISK QL MAX 1 Description The minimum percentage of disk fr
147. utilization percentage normalized with respect to the total system CPU time Appendix B Performance Manager Logical Names 405 PSPASDNS NAMES PSPASDNS NAMES Define this logical name in the Process table to a node name translation file specification Create this file to enable Real time data transport in DECnet Phase V environments when Node Synonyms are not defined The file contains translations from OpenVMS cluster to DECnet Phase V fullname The format of this ASCII file is one translation per line that consists of two names separated by a comma The first name is a one to six character OpenVMS cluster name and the second name is a one to five hundred eleven character DECnet Phase V fullname or segment thereof or address that DECnet OSI Phase V software will accept to establish a network connection For example LATOUR DEC TAY StanWilks PSPASEXAMPLES A system logical name defined by PSPASSTARTUP COM indicating the directory where Performance Manager example files are located This area may contain the following commands or rules PSPASDAILY COM Template command procedure to generate daily reports PSPASGETDATA COM Command procedure to create image from PSPASGETDATA MAR and C PSPASKB VPR Performance Manager factory rules source file PSPASGIVE DEVICE SERVICE For the disk statistics section of the Performance Evaluation Report the column labeled Busy is changed to Service Time when this logical nam
148. valid nodes for analysis For this reason the source of the data should always be chosen ahead of the classification of the data and the nodes to be analyzed If you select archived data for processing the history files will be locked blocking any archiving process Also if an archiving process is in progress the DECwindows interface will be suspended from reading the files until the archiving is complete 268 Performance Manager Administrator Guide Period to Be Analyzed How You Select Data for Analysis Classified By Specifies how the Performance Manager is to classify process activity in the Process Statistics Reports and in graphs presenting workload metrics Press MB1 on the Classify By option item and a menu appears The menu lists workload families Release on the family name you want The default option is None which results in the following summarizations All graph data will be displayed in the workload Other All Process Statistics reports will use the processing modes of Interactive Batch Network or Detached Enter any of the following information Start time End time Hold MB1 down on any of the date and time fields to see all available choices Release MB1 on the desired value Press and hold MB2 to advance through the possible values Press and hold MB3 to move back through the values These controls are desensitized if you have enabled the calendar option and have loaded a file of date ranges Press MB
149. value representing severe working set decrementing on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP_ACCESS_LOCK_RATEMET_F_ACCLCK Average XQP access lock rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP_ACCESS_LOCK_WAIT_RATEMET_F_XQPCACHEWAIT Average XQP access lock wait rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP_CACHE_HIT_RATEMET_F_HIT Average XQP cache hit rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP_CACHE_HIT_RATIO Derived This contains a value representing the ratio of disk cache hits to misses on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP XQP_CACHE_MISSEDIO_RATE Derived This contains a value representing the disk cache miss rate on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 532 Performance Manager Administrator Guide Performance Manager Data Cells XQP VOL AND DIR LOCK WAIT RATEMET F SYNCHWAIT Average XQP directory and volume synchronization lock wait rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL
150. value representing the count of images seen by the data collector in a nonpaged dynamic memory resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWPAG Derived This contains a value representing the count of images seen by the data collector in a paged dynamic memory resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO_IMAGES_SEEN_IN_RWPFF Derived This contains a value representing the count of images seen by the data collector in a paging file resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWQUO Derived This contains a value representing the count of images seen by the data collector in a job quota resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 471 Performance Manager Data Cells NO IMAGES SEEN IN RWSCS Derived This contains a value representing the count of images seen by the data collector in an SCS resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWSWP Derived This contains a value representing the count of images seen by the
151. was specified during the report time period divided by the number of its activations unless the number of activations is zero Total number of image activations due to interactive batch or network processes calculated for the report time period This example shows 197 interactive image activations Percentage of all direct I O due to interactive batch or network processes during the report time period In this example 24 70 percent of all direct I O was due to interactive processes Percentage of all buffered I O due to interactive batch or network processes during the report time period In this example 29 94 percent of all buffered I O was due to interactive processes Percentage of all CPU time used by interactive batch or network processes during the report time period In this example interactive processes consumed 12 30 percent of all CPU time Process type interactive batch network or detached or name of workload when primary key options are used or CLASSIFY BY Average number of active processes In this example there is an average of 32 active interactive processes during the 30 intervals Average number of inactive processes In this example there is an average of 53 inactive interactive processes during the 30 intervals Average number of private and global pages in the process s working set for the active processes Average number of soft page faults calculated by dividing the total number of soft pa
152. you perform the following actions Data to Be Analyzed Choose type and classification of data to be analyzed Set the beginning and ending date and time of the reporting period Set hourly schedule within the beginning and ending date Specify a Calendar file indicating specific dates for analysis Choose processing and report options Choose nodes for which data is to be reported Enter any of the following information Data Specifies the source of data to be analyzed Press MB1 on the Data option item and a menu appears The menu lists daily data collection definitions in your local schedule file and history file definitions in your local parameter file Release on the item you want The option menu disappears The menu item you chose is now the current source for performance data The default data source is daily data from the CPD collection definition Note Changing the source of data may change how it can be classified and which nodes can be analyzed If you have chosen a history file descriptor the Classify By options are changed to reflect those specified by the descriptor s definition If you choose daily data Classify By options reflect all workload families currently defined in the parameters file If a data source selection nullifies the current Classify By selection the Classify By selection is reset to the default of None For daily collection definitions the nodes specified by the collection definition are the only
153. you specify the secondary key as USERNAMES this column displays the total number of soft page faults for all images divided by the total number of image activations for the user 4 Number of hard page faults incurred by an image during the report time period divided by the number of activations If you specify the secondary key as USERNAMES this column displays the total number of hard page faults for all images divided by the total number of image activations for the user invoked per user 5 Average number of process private pages plus the global pages for this image or user during the report time period 6 Percentage of all direct I O attributable to an image or user during the report time period 7 Percentage of all buffered I O attributable to an image or user during the report time period 36 Performance Manager Administrator Guide 10 11 12 13 14 15 16 17 18 19 20 21 Performance Evaluation Report Percentage of all CPU time attributable to an image or user during the report time period Total elapsed time wall clock in seconds of an image or user divided by the number of its activations If the number of image activations is zero this measurement represents the total residence time of all activations of the image or all images if the USERNAMES option was specified Total amount of CPU seconds used by processes running an image or by a user if the USERNAMES option
154. 0 4046 22 MAR 2006 13 15 11 1101 10 2576 3 4636 2 3228 1 2826 0 5255 22 MAR 2006 13 30 28 1334 27 2356 12 2585 2 3256 1 2172 0 3333 22 MAR 2006 13 45 29 9854 29 1150 14 0548 2 1800 0 9782 0 1727 Chapter 4 Generate Historical Graphs 121 Format Graphs and Pie Charts PostScript Formatted Pie Chart Output illustrates PostScript formatted output for a pie chart It was generated with the following command ADVISE PERFORMANCE PIE CHART BEGINNING 22 MAR 2006 12 00 _ ENDING 22 MAR 2006 14 00 _ NODE BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY FORMAT POSTSCRIPT _ TYPE CUSTOM SYSTEM METRIC CPU TOTAL COMPOSITE CUSTOM GRAPH ae be whole pie reprereste tal Bes The following example Tabular Formatted Pie Chart Output illustrates tabular formatted output for a pie chart It was generated with the following command ADVISE PERFORMANCE PIE CHART BEGINNING 22 MAR 2006 12 00 _ ENDING 22 MAR 2006 14 00 _ NODE BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY FORMAT TABULAR _ TYPE CUSTOM SYSTEM METRIC CPU TOTAL COMPOSITE 122 Performance Manager Administrator Guide Format Graphs and Pie Charts PSPA CUSTOM GRAPH Nodes BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY Date 22 MAR 2006 12 00 14 00 The whole pie represents 11 79 CPU Total Busy BUGDEV 5 68626 BYOB 3 59973 CINAMN 0 96963 ORIPAS 0 49438 BATCH CHATTY MIGHTB 0 18164 The following example CSV
155. 1 15 JAN 2008 13 00 0 0092 0 005 0 0 0 0 15 JAN 2008 13 02 0 0092 0 005 0 0 0 0 15 JAN 2008 13 04 0 0075 0 0017 0 0 0 0 15 JAN 2008 13 06 0 1425 0 1233 0 1067 0 0425 0 0292 0 015 15 JAN 2008 13 08 0 0075 0 0017 0 0 0 0 Resulting File After Merge and Edits PSPA CUSTOM GRAPH Node RX8620 Date 15 JAN 2008 13 00 13 10 Metric Values are Stacked eg Added to the left Units PROCESSOR Total Busy Time RX8620 0 RX8620 1 RX8620 2 RX8620 3 RX8620 4 RX8620 5 RX8620 6 RX8620 7 RX8620 8 RX8620 9 RX8620 10 RX8620 11 15 JAN 2008 13 00 0 9183 0 2475 0 1250 0 1250 0 1108 0 0033 0 0092 0 005 0 0 0 0 15 JAN 2008 13 02 0 9192 0 2450 0 1233 0 1233 0 1100 0 0025 0 0092 0 005 0 0 0 0 15 JAN 2008 13 04 0 9192 0 2458 0 1250 0 1250 0 1058 0 0008 0 0075 0 0017 0 0 0 0 15 JAN 2008 13 06 1 1150 0 3717 0 2358 0 2208 0 1892 0 0175 0 1425 0 1233 0 1067 0 0425 0 0292 0 015 15 JAN 2008 13 08 0 9175 0 2433 0 1242 0 1242 0 1067 0 0000 0 0075 0 0017 0 0 0 0 Appendix F How You Graph Seven or More CPUs 561 Step 4 Send the CSV File to a Windows Machine Step 4 Send the CSV File to a Windows Machine Send the CSV file to a Windows machine Utilize whatever method works best for you FTP for example Step 5 Create the Graph in Excel To create the graph in Excel 1 2 Open the CSV file in Excel Select the data beginning with cell A6 through the lower right cell you
156. 1 on the Set time option button and an option menu appears Enter any of the following options m Default today Specifies midnight to now m Yesterday Specifies yesterday from midnight to midnight m Most recent hour Specifies the last 60 minutes m First hour of today Specifies the time period of 00 00 to 01 00 m Advance by a day Increment the beginning and end dates by one m Backup by a day Decrement beginning and end dates by one m Specify text Release on the Specify text menu item to specify the beginning and end day and time from the keyboard Chapter 8 Use the DECwindows Motif Interface 269 How You Select Data for Analysis Performance Manager displays the Performance Manager Time Selection Box Performance Time Selection Box z Specify Start and End time as DD MMM YYYY HH MM Start Time 20 JAN 2006 0800 End Time 20 Jan 2006 1000 ox Reset Cancel Help The day and time can be entered in the format shown in the window Clicking on the OK button applies the start and end times and removes the dialog box from the screen The Reset button restores the start and end times to those displayed in the Performance Manager Data Selection box Clicking on the Cancel button removes the dialog box without changing the time fields currently displayed in the Data Selection box Schedule Specifies a subset of hours within your beginning and ending reporting period Release on the Modify button a
157. 1992 00 00 pana files 7 days anor Collection interval 120 seconds Stop date 01 3ax 2010 00 00 Enter the minimum average queue size that a disk must have to start collecting the hottest files for that disk This value is a decimal number less than 100 00 and greater than or equal to 0 00 The default is 0 33 If the toggle button is off no hot file data is collected If the toggle button is on you must specify a queue length or accept the default value Collection interval The Collection interval specifies the number in seconds in a sampling interval At the end of this time period data is recorded into an interval record The CPD collection definition has an interval of 120 seconds which cannot be modified You can specify an interval from 1 to 3600 seconds for other definitions Delete files after Enter the number of days that data is to be retained The default is seven days for the primary data collector Data files are automatically deleted from the database when they are older than the specified number of days For alternate data collectors the default is 99 999 days which lets you control the how long data files are retained This value should be coordinated with any data archiving and the amount of free space required for the database area For information on archiving see the Performance Agent Administrator Guide Start date The date and time on which the collection definition goes into effect A data collect
158. 2 Click YES to load and view the parameters file Write the Parameter File To write the parameter file m Release on the Write Parameter File menu item if you want to save changes you have made to the file OR If you have not made changes this option is desensitized Load the Binary Graph Data The Performance Manager software lets the saving of analyzed data required to support graphing functions in a summarized format To load one of these files and use the graphing options release on the Load Binary Graph Data menu item The Performance Manager software displays the Load Graph Data Dialog Box a Load Graph Data Dialog Box Load Graph Data from binary loadable file r i il OK Cancel 264 Performance Manager Administrator Guide How You Control the DECwindows Interface To load the binary graph data m Specify a file to which a selection of previously saved graphing data is to be loaded If you previously requested data analysis using an option from the Select menu access to that analysis data is lost Normally Performance Manager data is selected and processed prior to performing any of the display functions available This selection process can be very time consuming especially if you have activated all processing options If at a later date you want to view other graphs or pie charts associated with the original selection you might want to use the SAVE and LOAD graph data features Otherwise the
159. 2 To restore the menu click MB3 and a pop up menu displays 3 Click the Restore Panel Menu option If no instrument is active when there is no panel menu the pop up menu options are limited to the following actions m Close m Save m Restore Panel Menu 370 Performance Manager Administrator Guide How You Set the Panel Options Save the Panel To save a panel 1 Pull down the File menu and release on the Save menu item 2 To specify a new panel name click the Save As menu item The Panel Name dialog appears 3 Enter the new panel name and click the OK button Close the Panel To close an instrument pane 1 Pull down the File menu and click the Close menu item If you modified the instrument Performance Manager displays the following message 2 Click Yes to save the modifications and close OR Click No to close without saving your changes Chapter 10 Customize the DECwindows Motif Real time Display 371 Chapter 11 Use the Character Cell Real time Display This chapter provides information about the Performance Manager character cell Real time Display This section contains the following topics Character Cell Display Functions see page 373 Start the Character Cell Displays see page 374 Control the Displays see page 374 Display Multi node Statistics see page 376 Display Single Node Statistics see page 378 Display Process Information see page 381 Display Disk Information see pag
160. 7 Motif interface e 257 260 controlling e 260 data collector status e 259 data selected for processing e 259 files locked e 259 main window e 258 quitting e 266 read parameter file e 264 saving reports e 260 work in progress e 261 write parameter file e 264 Motif Real time display e 321 commands menu e 324 controlling e 322 CPU modes panel e 326 CPU queue panel e 329 CPU utilization panels e 325 default panel e 321 default panels e 324 Disk info panel e 336 Disk rate panels e 335 Disks panel e 335 Hard fault rate panels e 330 launching panels e 323 Memory allocation panel e 333 navigate panels e 323 Page faults panel e 330 Page file allocation panel e 334 panel hierarchy e 324 panel instrument e 341 panel manager e 322 Process wait panel e 329 starting e 321 341 system overview panel e 324 User faults panel e 332 User memory panel e 334 DECwindows Motif interface starting e 257 Default ADVISE PERFORMANCE GRAPH BEGINNING e 191 ENDING e 191 ReGIS output file type e 191 SCHEDULE e 191 STACK e 191 Y AXIS MAXIMUM e 191 245 ADVISE GRAPH ANSI TABULAR output file type e 191 FORMAT overrides e 191 X POINTS e 191 data selection of e 266 domain for rule element e 150 graph type e 108 191 number of hot files 61 panels navigating within e 323 rules file type e 176 threshold values e 137 X POINTS value for e 123 Delete panel instruments e 342 DEQ e 84 Dir FCB e 82 Direct I Os e 81 Disk device metrics for
161. 70438 freep size of the free page list 158 7 6 5 net proc of network processes 167 7 2 0 bat_proc of batch processes 176 7 122 0 other_proc of other types of 185 7 processes inter_proc of interactive 194 7 processes 219 7 total_proc total process count 203 7 0 0 como state of procs in comp 212 7 outswap state 217 7 balance set of procs in the balance 221 7 set 0 0 inswp count of inswap operations 230 7 0 0 outswp coun of outswap operations 239 7 t 0 0 hdr_inswp of header inswap 248 7 operations 0 0 hdr_outswp of header outswap 257 7 operations 542 Performance Manager Administrator Guide CPU Statistics Data Record CPU Statistics Data Record The following example commands show how to export and display CPU statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP CPU CLASS NODEFAULT CPU _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP CPU ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA CPU 56 8 0 9 0 3 0 0 1 3 0 0 40 4 58 0 1 3 2315 0 34 4 0 1 4 8 35 1 9 8 0 0 9 8 42 9 22 0 12 4 22 7 0 15 1 0 3 Example Data item Description Position Length 56 8 interrupt of time in interrupt 91 7 mode 0 9 kernel of time in kernel 100 7 mode 0 3 exec of time in executive 109 7 mode 0 0 super of time in supervisor 118 7 mode 1 3 user 9
162. 9 set archiving e 317 workload families e 317 file where defined e 191 Hot file queue size of e 298 I O rule e 26 Image CPU panel e 329 name e 90 309 Informational message e 403 Instruments connect e 324 disconnect e 324 setting interval of e 324 InSWP e 77 Inter stack e 77 Interactive displays in Command mode e 239 Interval record frequency of e 156 time stamped e 159 value history file descriptors e 317 INVESTIGATE keypad e 392 J JOBS graph type e 191 K Kernel e 77 Key settings primary keys e 34 secondary keys e 34 Keywords NO STACK e 191 245 ANSI e 191 COLOR e 117 HEIGHT e 191 LINE e 117 230 PATTERN e 117 ReGIS e 117 191 582 Performance Manager Administrator Guide STACK e 245 TABULAR e 191 THRESHOLD e 245 TITLE e 191 245 TODAY e 191 TOMORROW e 191 WIDTH e 191 X POINTS e 191 Y AXIS MAXIMUM e 191 245 YESTERDAY e 191 Knowledge base e 15 auxiliary build e 135 172 use e 135 definition e 176 L Launch panels assign e 362 Launching panels e 323 LINE keyword e 117 230 Lock rates e 84 LOCKS graph type e 191 Logical names e 405 Lognam trans e 81 Lost CPU e 78 Low threshold bar graph setting of e 347 strip chart setting of e 348 M Mailbx reads e 81 writes e 81 Maximum data value bar graph setting of e 347 strip chart setting of e 348 Mem queue e 76 Memory allocation panel e 333 panel e 333 MEMORY UTITLIZATION graph type e 191 Messages e 403 Meter parts specify e 356
163. ADVISE PERFORMANCE COMPILE If the INTERFACE qualifier is used to start the DECwindows interface refer to the chapter Use the DECwindows Motif Interface see page 257 for more information The following table lists the ADVISE PERFORMANCE command options Option Function COMPILE Compiles user rules DISPLAY Activates the Performance Manager Real time Display Interface EXPORT Activates the data export facility GRAPH PIE CHART Activates the Performance Manager graphing facility REPORT Activates the Performance Manager reporting facility SHOW VERSION Identifies the current version of the Performance Manager module ADVISE PERFORMANCE COMPILE Format Parameter Description The ADVISE PERFORMANCE COMPILE command invokes the Performance Manager rules compiler to compile a set of user rules from the specified file ADVISE PERFORMANCE COMPILE file spec file spec The file specification of a text file that contains user defined rules or disabled Performance Manager factory rules The default file type is VPR ADVISE PERFORMANCE COMPILE invokes the rules compiler to compile a set of rules into the auxiliary knowledge base The output file is a compiled version of the rules and thresholds in an efficient format used to produce an Analysis Report To use the auxiliary knowledge base specify the file name with the RULES qualifier when you generate either a Brief Analysis or an Analysis report 176 Performance Man
164. AGE qualifier is used Chapter 7 Use Command Mode Commands 241 SELECT BEGINNING date Specifies the beginning time for the data selection By default 00 00 is used CLASSIFY BY USERGROUP family name Specifies the family name which dictates how to classify the workload for workload graphs and the process statistics section of the Performance Evaluation Report By default no classification is used COLLECTION DEFINITION collection definition name Specifies the name of the Collection Definition and hence the collected data that you desire to use for graphs and reports If you omit this qualifier daily data is obtained from the Collection Definition called CPD To view the Collection Definitions that you have available use the DCL command ADVISE COLLECT SHOW ALL If you want to use history data instead of daily data use the HISTORY qualifier instead of the COLLECTION DEFINITION qualifier These two qualifiers are mutually exclusive DATES filespe Specifies that a file containing a series of date ranges is to be used in place of the BEGINNING and ENDING qualifiers Each line in the dates file should look like the following code dd mmm yyyy hh mm ss cc dd mmm yyyy hh mm ss cc The time may be omitted entirely or may be truncated Any truncated parts of the time are defaulted to 0 The periods of time represented by each line in the file need not be contiguous but they must be in ascending order DATES i
165. AP SYSGEN IRPCOUNTPAR F IRPCOUNT The value of the SYSGEN parameter IRPCOUNT which sets the number of pre allocated intermediate request packets for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN IRPCOUNTVPAR F IRPCOUNTV The value of the SYSGEN parameter IRPCOUNTV which is the virtual IRP count which is the number of intermediate request packets to which the IRP list may be extended for the current interval for the local node Data Type NUMERIC Appendix C Performance Manager Data Cells 507 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LCKMGR MODEPAR F LCKMGR MODE The value of the SYSGEN parameter LCKMGR MODE which controls the use of the Dedicated CPU Lock Manager for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LOAD SYS IMAGEPSPAR F LOAD SYS IMAGES The value of the SYSGEN parameter LOAD SYS IMAGES which controls the loading of system images described in the system image data file VMS SYSTEM IMAGES for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC
166. APH a cede cde aha ee cde cr Cu cede cl ccr a ba bp CEU reus 245 PIEZCHARTI ore is ee E AA ee ee ee UM 248 REPOR e d 249 s te ama ATAT AN ee ee ee ee ee ee ee ee 255 SPAWN 6 344454444400540454445406164344404003404043639040030040404944404003405340640005400048 255 40 256 Execute cla LA 256 Chapter 8 Use the DECwindows Motif Interface 257 Start the DECwindows Motif Interface 2 sla enn 257 Use the Main WINDOW 2 5608 da cuore dot at aw ete emt hah a aa ial ring 258 Main Window Status Information sss rn 259 How You Control the DECwindows Interface 2 ssl 260 Contents 7 Save the Reports 02405 dsl AN E A eR a s E se ADA E NANA MANA Wa TE Ee 261 Monitor the Work in Progress 22 2 sss 263 Read the Parameter File Nka aakuin 264 Write the Parameter File sr 264 Load the Binary Graph Data 2222 2200 sse 264 Save the Binary Graph Data 2 22 2220 esr 265 Quit the Sessiori 222225226 i iia a Gad d aga dad Paggdquddqudqqudqquadaqugqqugqqugsdqu 266 How You Select Data for Analysis 2 sss rn 266 Select Today s Data on oe Hanapan Deve vod Dove KATA BAY Nah Da Dee pd pep EEIT 266 Select Specific Data 2 esee 267 Select the Last HOUr 2 2 6 0 6 bee e bl ae be BNG PAP rra A EVE 277 Use Custom Default Settings ss rn 277 How You Display Analyzed Data sussslssllsssllss sr es 278 Brief Analysis Report 2 278 Full Analysis Report
167. ASKB VPR Family name is an identifier for a group of workload definitions also known as a workload transaction or usergroup family There are two family types Usergroup families contain workload definitions based on user criteria Transaction families contain workload definitions based on image and process data Specify either family with the CLASSIFY BY qualifier to control use of the family for the reporting facility A file type is specified by one of the following extensions in its name COM Various command files CPD Cluster Performance Data file DAT Parameters and schedule file EXE Various image files HLB Parameter Editor help file KB Compiled rules Knowledge Base file LIB Model library file LIS A report file LOG Data collection process error log file MAR Sample macro application file MDL Model file name History file name is history file descriptor also alternate data file name name JOU History journal file REG A ReGIS graph written to a disk file TXT Holidays and message files VPR Performance Rules source file Granularity is a Performance Manager parameter file element that specifies for each history file descriptor how often a history file is created A hard page fault occurs each time a process references a virtual page that is not in its working set and requires a read operation from disk a hard page fault is generated 568 Performance Manager Ad
168. ATION CPU DISK FILE or TAPE domains with a process domain index specifier The Index Specifier Data Cell section describes Index Specifier data cells The SUMMARY and CLUSTER domains do not directly correlate to Performance Manager data records however they are derived from them Chapter 5 Customize the Knowledge Base 157 Data Cell Types and Use The data cells available in the SUMMARY domain are metrics that are maximums or averaged from all of the interval records After processing all the data for a single node the summary data cells become available in the SUMMARY domain The last parameter sub record is also available in the SUMMARY domain The number of times the Performance Manager tests a rule in the SUMMARY domain is equal to the number of nodes processed After the Performance Manager processes all nodes data the data cells available in the CLUSTER domain are disk statistics that represent the cluster perspective of the I O traffic The Performance Manager combines data from each node s two minute disk sub records into a set of two minute cluster records in memory The Performance Manager provides items such as cluster wide throughput and operation rates however there is no longer any association with the current two minute data record Metric and parameter sub record values available from the LOCAL domain are not accessible from the CLUSTER domain The following table lists the seven types of data cells Data Cell
169. BT 11 47 34 264 2 24 2 76 0 02 4 0 08 BACKUP 6 85 68 454 21 04 1 95 0 02 6 0 18 CAPOLYSCASEN 1 278 33 574 0 08 0 07 0 00 0 0 03 CAPOLYSFORCE 2 56 17 209 0 05 0 08 0 00 8 0 02 CAPOLYSTRAP 18 264 33 293 1 22 1 10 0 01 1 0 03 cou 4 177 21 443 0 53 0 48 0 00 0 0 05 CONVERT 2 31 32 336 0 08 0 06 0 00 0 0 02 copy 9 51 8 307 0 52 0 48 0 00 1 0 03 CREATE 9 22 M 354 0 11 0 80 0 00 0 0 01 4 Traditional Focus 4 xj dj INTERACTIVE 164YME 2 gt To navigate through the Process Statistics Reports click the Traditional Focus label An option menu appears listing all of the Process Statistics Reports Release on the report type you want displayed You can also advance through the reports by clicking on the appropriate direction arrows The Section label here labeled INTERACTIVE changes according to the report section being viewed and will be desensitized if irrelevant The arrows on either side of this label allow you to advance to other sections of the report Holding MB1 down on the section label will display all other available reports Releasing MB1 on the desired report name will update the Detail window Press on the Node button and an option menu appears listing the nodes on which the report is based Release on the node you want displayed 282 Performance Manager Administrator Guide How You Display Analyzed Data To exit from the Process Statistics Window pull down the File menu and release on Close
170. C Channel Thruput e 191 HSC Disk IO e 191 HSC Disk Thruput e 191 HSC IO e 191 HSC Tape IO e 191 HSC Tape Thruput e 191 HSC Thruput e 191 Image Activations Users e 191 Image Activations Workload e 191 Image I O Operations e 191 Images Activated e 191 Images Volume IO e 191 Interrupt Stack Procesor e 191 IO Rule Occurrences e 191 KB Sent Rate e 191 Kernel Mode Procesor e 191 Kilobyte Mapped Rate e 191 Kilobyte Received Rate e 191 Memory Rule Occurrences e 191 Messages Received e 191 Messages Sent e 191 MP Synch Mode Procesor e 191 MSCP I O Operations files e 191 Operations Disk Device e 191 Operations Disk Volume e 191 Operations Files e 191 Percent Freeblks Disk Device e 191 Percent Freeblks Disk Volume e 191 Percent Usedblks Disk Device e 191 Percent Usedblks Disk Volume e 191 PG amp SWP Operations Disk Device e 191 PG amp SWP Operations Disk Volume e 191 PG amp SWP Operations Files e 191 Pool Rule Occurrences e 191 Queue Disk Device e 191 Queue Disk Volume e 191 Queue HSC Channel e 191 Read Operations Disk Device e 191 Read Operations Disk Volume e 191 Read Operations Files e 191 Resource Rule Occurrences e 191 Response Time Disk Device e 191 Response Time Files e 191 Response time images e 191 Response time users e 191 Response time workloads e 191 Rule Occurrences e 191 Service Time Disk Volume e 191 Split Operations Disk Device e 191 Split Operations Disk Volume e 191 Split Operations Files e 191 Supervi
171. CA Performance Management for OpenVMS Performance Manager Administrator Guide r 1 Ca This documentation and any related computer software help programs hereinafter referred to as the Documentation is for the end user s informational purposes only and is subject to change or withdrawal by CA at any time This Documentation may not be copied transferred reproduced disclosed modified or duplicated in whole or in part without the prior written consent of CA This Documentation is confidential and proprietary information of CA and protected by the copyright laws of the United States and international treaties Notwithstanding the foregoing licensed users may print a reasonable number of copies of the documentation for their own internal use and may make one copy of the related software as reasonably required for back up and disaster recovery purposes provided that all CA copyright notices and legends are affixed to each reproduced copy Only authorized employees consultants or agents of the user who are bound by the provisions of the license for the product are permitted to have access to such copies The right to print copies of the documentation and to make a copy of the related software is limited to the period during which the applicable license for the Product remains in full force and effect Should the license terminate for any reason it shall be the user s responsibility to certify in writing to CA that all copies and partial
172. CC NO TOP CPU WORKLOADS NO TOP DGS D NO TOP DGS S NO TOP DIRIO USERS NO TOP DISKIO IMAGES NO TOP DISKIO WORKLOADS NO TOP FAULTING IMAGES NO TOP FAULTING WORKLOADS NO TOP FREEBLK VOLUMES NO COMPUTE QUEUE NO CPU UTILIZATION NO DECNET NO FAULTS NO JOBS NO MEMORY UTILIZATION NO RESPONSE TIME NO TOP BDT W NO TOP BLKS S NO TOP BUFIO USERS NO TOP BUSY DISKS NO TOP BUSY VOLUMES NO TOP CHANNEL QUELEN NO TOP CLUSTER RULE OCC NO TOP CPU IMAGES NO TOP CPU USERS NO TOP CR W NO TOP DGS R NO TOP DIRIO IMAGES NO TOP DIRIO WORKLOADS NO TOP DISKIO USERS NO TOP EXEC PROCESSOR NO TOP FAULTING USERS NO TOP FREEBLK DISKS NO TOP HARDFAULTING IMAGE S Chapter 6 Performance Manager Commands 201 ADVISE PERFORMANCE GRAPH NO TOP HARDFAULTING USERS NO TOP HSC DISK IO NO TOP HSC IO NO TOP HSC TAPE THRUPUT NO TOP IDLE PROCESSOR NO TOP IMAGE VOLUME IO NO TOP IOSIZE DISKS NO TOP IOSIZE IMAGES NO TOP IOSIZE WORKLOADS NO TOP IO FILES NO TOP IO VOLUMES NO TOP KB RC NO TOP KERNEL PROCESSOR NO TOP MGS R NO TOP MP SYNCH PROCESSOR NO TOP PAGING DISKS NO TOP PAGING VOLUMES NO TOP PRCT FREE DISKS NO TOP PRCT USED DISKS NO TOP QUEUE DISKS NO TOP READ DISKS NO TOP READ VOLUMES NO TOP RESIDENT USERS NO TOP RESOURCE RULE OCC NO TOP RESPONSE TIME FILES NO TOP RESPONSE TIME USERS NO TOP RESPONSE TIME WORKLOADS NO TOP SPLITIO DISKS NO TOP SPLITIO VOLUME
173. CFG CPU DSK FIL LOC PRO TAP Target Domains CPU CPU_INTERRUPTCPU_F_INTERRUPT This contains the value representing the percent of time spent on the interrupt stack for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU Appendix C Performance Manager Data Cells 425 Performance Manager Data Cells CPU INTERRUPT TALLY Derived This contains the value representing the percent of time spent on the interrupt stack for the physical CPU represented by the current CPU subrecord and interval Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU CPU IS PRIMARYCPU C PRIMID This contains either the values 1 if the CPU for the current CPU subrecord is the primary CPU or a 0 if it is not the primary CPU Data Type NUMERIC Domains CPU CPU IS RUNNINGCPU C RUN This contains either the values 1 if the physical CPU for the current CPU subrecord is running or a O if it is stopped Data Type NUMERIC Domains CPU CPU KERNELCPU F KERNEL This contains the value representing the percent of time spent in KERNEL mode for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU KERNEL TALLY Derived This contains the sum of the values representing the percent of time in KERNEL mode for the current CPU subrecords which were selected by the most recent CPU SCAN routine operation Data Type
174. CHED COMQ DETACHED PROCESSES DIRIO DISK PAGING DISK SWAPPING ADVISE PERFORMANCE GRAPH Description Percent CPU time used by network jobs Percent CPU time for which the Performance Manager did not capture process data Percent CPU time not in idle mode The VUP rating of the CPU Percentage of CPU SWAPPER busy Percentage of CPU Multi I O busy Percentage of CPU Any I O busy Percentage of CPU idle page wait Percentage of CPU idle swap wait Percentage of CPU idle page or swap wait percentage of CPU and I O idle Percentage of CPU only busy Percentage of I O only busy Percentage of CPU and I O busy Average number of processes in currently executing process state Number of deadlocks found by OpenVMS per second Number of deadlock searches per second Departing local packets per second Number of computable detached processes Number of detached processes Direct I O per second Number of paging I O operations per second Number of swapping I O operations per second Chapter 6 Performance Manager Commands 211 ADVISE PERFORMANCE GRAPH Channel Description DISK USER Number of user I O disk operations per second DZROFAULTS Number of demand zero page faults per second ERASE QIO Number of Erase QIO operations per second EXEC Percent CPU time charged to executive mode FILE OPEN Number of files opened per second FILE SYS Percent CPU time spent in the file system System Description FP
175. CONFIGURATION or TAPE domains Specify the PROCESS domain data cell with the index specifier as a parameter A rule expression for a rule in the LOCAL domain is as follows PROCESS CPUTIME TOP CPU PROC X This expression calculates which process has the highest CPU time Although PROCESS CPUTIME is a PROCESS domain data cell TOP CPU PROC X is a LOCAL domain index specifier that has a target domain of PROCESS So you can reference any PROCESS domain data cell from a rule in the LOCAL domain by using an index specifier with the target domain of PROCESS The maximum index depth for index specifiers is two Implement Changes This section provides scenarios for customizing the knowledge base with the following actions m Disabling an existing rule m Modifying an existing rule Adding a new rule Changing a threshold value Changing a rule literal value While the scenarios assume that your auxiliary rules file is called MYRULES VPR you can name it whatever you like provided that it conforms to standard OpenVMS naming conventions Chapter 5 Customize the Knowledge Base 163 Implement Changes There are two files to which you probably need to refer when making changes to the factory rules The first is a copy of the source file for the factory rules located in PSPASEXAMPLES PSPASKB VPR The second is the message text file which contains the Conclusions and Evidence headings located in SYSSSYSTEM PSPASMSG TXT You should mak
176. Cwindows Motif Interface 309 How You Customize Either images or users Indicates that the Performance Manager will match either the image names or the user criteria of a process record to include the process data in the workload Both images and users Indicates that the Performance Manager must match both the image names and the user criteria of a process record to include the process data in the workload Transaction units This is an option menu with two entries Press MB1 on the current setting to view the choices Click either image termination or terminal responses to indicate how response time should be evaluated This will affect the workload frequency when building a model See the ADD WORKLOAD command in the CA Performance Management for OpenVMS Agent Administrator Guide for information about transaction units Images Click the text entry field to activate the text insertion cursor To add an Image name enter a name and press Return or click the Enter button The name will appear in the list box and be cleared from the text entry field If you wish to preserve lower case characters enclose image names in double quotes when you enter them To delete an Image name click the name in the image name list box then click the Delete button The name is removed from the list box To remove multiple names click all their names and then click the Delete button A list of image names can be provided through a file Use the at s
177. D DISKS graph type e 191 TOP PRCT USED VOLUMES graph type e 191 TOP QUEUE DISKS graph type e 191 TOP QUEUE VOLUMES graph type e 191 TOP READ DISKS graph type e 191 TOP READ FILES graph type e 191 TOP READS VOLUMES graph type e 191 TOP RESIDENT IMAGES graph type e 191 TOP RESIDENT USERS graph type e 191 TOP RESIDENT WORKLOADS graph type e 191 TOP RESOURCE RULE OCC graph type e 191 TOP RESPONSE TIME DISKS graph type e 191 TOP RESPONSE TIME FILES graph type e 191 TOP RESPONSE TIME IMAGES graph type e 191 TOP RESPONSE TIME USERS graph type e 191 TOP RESPONSE TIME WORKLOADS graph type e 191 590 Performance Manager Administrator Guide TOP RULE OCCURRENCES graph type e 191 TOP SERVICE VOLUMES graph type e 191 TOP SPLITIO DISKS graph type e 191 TOP SPLITIO FILES graph type e 191 TOP SPLITIO VOLUMES graph type e 191 TOP SUPER PROCESSOR graph type e 191 TOP TERMINAL INPUT IMAGES graph type e 191 TOP TERMINAL INPUT USERS graph type e 191 TOP TERMINAL INPUT WORKLOADS graph type e 191 TOP TERMINAL THRUPUT IMAGES graph type e 191 TOP TERMINAL THRUPUT USERS graph type e 191 TOP TERMINAL THRUPUT WORKLOADS graph type e 191 TOP THRUPUT DISKS graph type e 191 TOP THRUPUT FILES graph type e 191 TOP THRUPUT IMAGES graph type e 191 TOP THRUPUT USERS graph type e 191 TOP THRUPUT VOLUMES graph type e 191 TOP THRUPUT WORKLOADS graph type e 191 TOP USER IMAGE ACTIVATIONS graph type e 191 TOP USER PROCESSOR graph ty
178. D X GT DISK IO RATE DISK OVER QL THRESHOLD X 00 5 OCCURRENCES GE 4 EVIDENCE Volume w Highest Queue Length Stim tim file EXEC I0s sec Pag I0s sec Swp I0s sec sys mode Time of occurrence BRANDY1 1 0CT 09 04 00 BRANDY1 1 0CT 09 06 00 BRANDY1 1 0CT 09 08 00 BRANDY1 1 0CT 09 10 00 The following statements are keyed to the columns in the previous I O Rule Analysis report 1 Name of the volume on which excessive swapping occurred This volume had the highest queue length during the interval exampled 2 Average number of I Os per second to the volume from the node currently being analyzed SUPPLY 3 Number of paging I Os per second to the volume The value of paging I Os per second is a subset of the total I Os per second as described above 4 Number of swapping I Os per second to the volume This value is also a subset of the total I Os per second 5 Percentage of CPU time spent in the file system 6 Percentage of CPU time spent in executive mode 7 Beginning time of the interval in which the condition occurred 26 Performance Manager Administrator Guide Analysis Reports In the previous example I O Rule Analysis report the disk queue length on volume BRANDY1 exceeded its threshold on four occasions during the reporting interval In each occurrence less than 30 percent of the CPU time was spent in the file system and less than 20 percent of the CPU time was spent in executive mode Swapping I Os per
179. D for the current process subrecord on the local node for the current interval record Data Type NUMERIC Domains PRO MAXIMUM_DISK_QUEUE Derived This contains a value representing the maximum queue length for the current disk for all cluster members during the current interval Data Type NUMERIC Domains CLU 466 Performance Manager Administrator Guide Performance Manager Data Cells MAXIMUM IRPS INUSE Derived This contains a value representing the maximum number of IRPs in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY MAXIMUM LOCKS INUSE Derived This contains a value representing the maximum number of locks in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY MAXIMUM LRPS INUSE Derived A count of the maximum number of large request packs in use for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY MAXIMUM RESOURCES INUSE Derived This contains a value representing the maximum number of resources known by the local node for all of the intervals Data Type NUMERIC Domains SUMMARY MAXIMUM SRPS INUSE Derived This contains a value representing the maximum number of SRPs in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY MAXIMUM WORKING SET SIZE Derived This contains the value of the maximum working set size of all processes for the local node for the
180. DS Top Direct I O Workloads Plots the top five plus Other Users creators of direct I O by workload name The metric graphed is DIRIO TYPE TOP DISKIO IMAGES Top Image I O Operations Plots the top five plus Other Images creators of disk I O by image name The metric graphed is DSKIO TYPE TOP DISKIO USERS Top User Disk Operations Plots the top five plus Other Users creators of disk I O by user name The metric graphed is DSKIO TYPE TOP DISKIO WORKLOADS Top Workload Disk Operations Plots the top five plus Other creators of disk I O by workload name The metric graphed is DSKIO TYPE TOP EXEC PROCESSOR Top Exec Mode Processor Plots the five plus Other processors in terms of time spent in executive mode as a percent of CPU time The metric graphed is P EXEC TYPE TOP FAULTING IMAGES Top Faulting Images Plots the top five plus Other Images creators of page faults by image name The metric graphed is FAULTS TYPE TOP FAULTING USERS Top Faulting Users Plots the top five plus Other Users creators of page faults by user name The metric graphed is FAULTS TYPE TOP FAULTING WORKLOADS Top Faulting Workloads Plots the top five plus Other Users creators of page faults by workload name The metric graphed is FAULTS TYPE TOP FREEBLK DISKS Top Freeblks Disk Device Plots the top five plus Other disk devices in terms of number of free disk pages The metric graphed
181. Domains COM CFG CPU DSK FIL LOC PRO SUM TAP Appendix C Performance Manager Data Cells 511 Performance Manager Data Cells SYSGEN MULTITHREADPAR F MULTITHREAD The value of the SYSGEN parameter MULTITHREAD which controls the availability of kernel threads functions for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN NPAGEDYNPAR F NPAGEDYN The value of the SYSGEN parameter NPAGEDYN which sets the size of nonpaged dynamic pool in bytes for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN NPAGEVIRPAR F NPAGEVIR The value of the SYSGEN parameter NPAGEVIR which defines the maximum size to which NPAGEDYN can be increased for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PAGEDYNPAR F PAGEDYN The value of the SYSGEN parameter PAGEDYN which sets the size of the paged dynamic pool in bytes for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC
182. Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LOCKDIRWTPAR F LOCKDIRWT The value of the SYSGEN parameter LOCKDIRWT which is the lock manager directory system weight which determines the portion of the lock manager directory which will be handled by this system for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LOCKIDTBLPAR F LOCKIDTBL The value of the SYSGEN parameter LOCKIDTBL which sets the initial number of entries in the system Lock ID table and defines the amount by which the Lock ID table is extended whenever the system runs out of locks for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 508 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN LONGWAITPAR F LONGWAIT The value of the SYSGEN parameter LONGWAIT that defines how much real time in seconds must elapse before the swapper considers a process to be temporarily idle This applies to the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LRPCOUNTPAR F LRPCOUNT The value of the SYSGEN parameter LRPCOUNT which sets the number of pre allocated large reques
183. Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG DATAGRAMS_SEND_RATESCS_F_DGSENT This contains the value representing the number of datagrams sent per second from the local node to the remote node for the current configuration record and interval Data Type NUMERIC Domains CFG DATAGRAMS_SEND_TALLY Derived This contains the sum of the values representing the number of datagrams sent per second from the local node to the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG DEADLOCK_FIND_RATEMET_F_DLCKFND This contains the number of deadlock finds per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DEADLOCK_SEARCH_RATEMET_F_DLCKSRCH This contains the number of deadlock searches per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 43 Performance Manager Data Cells DECNET RECV BUFF FAIL RATEMET F RCVBUFFL This contains the number of times per second the DECNET receiver buffer failed for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP DECNET TRANSIT CONGSN LOSS RATEMET F TRCNGLOS Average DECNET transit congestion loss
184. E ESE EE EE E EEE 160 Index Specifier Data Cell 2 2 162 Implement Changes iissssse a a hah ha hr ahh ah rh has 163 Disable an Existing Rule 222 2 ra 164 Modify an Existing Rule ise esed es eG PR OG due d i qu daa dae d pepe 164 Add a New Rule 2 use eere rr err eer rre rer rrr e eR 166 Change a Threshold Value EG cence nee cea n seen nee EEEE 167 Change a Rule Literal Value 22 2 t nets 170 Build an Auxiliary Knowledge Base 2 22 22 tee ett ls 172 Use an Auxiliary Knowledge Base for Reporting and Archiving 2 eee eee 173 Chapter 6 Performance Manager Commands 175 ADVISE PERFORMANCE 2 020c0acaeae 444402404005 eae eeee NA weed eos REEE ese evade l4 eie RES 175 ADVISE PERFORMANCE COMPILE 2 22222220 000 cece sess 176 ADVISE PERFORMANCE DISPLAY 3 222 020 s0 s 00s000es0sesususuieavavassesesesetesuiuieeeaaae 178 ADVISE PERFORMANCE EXPORT 0 00 0c cece etn ras 182 ADVISE PERFORMANCE GRAPH 0 0 00 ccc cee ssa 191 ADVISE PERFORMANCE PIE CHART kae ue e eee pere eee eee dee eae eee 232 ADVISE PERFORMANCE REPORT 22222222 esser 234 ADVISE PERFORMANCE SHOW VERSION sssssssslsls essa 238 Chapter 7 Use Command Mode Commands 239 ADVISE PERFORMANCE lt s 22 703 na na am 0400 ucudu mdacR4 icidixmddamici4dasSdcacac imidanagamivayio 239 SELECT 240 LOAD aaa Na paa Eq Ee d deg a e P EP E nere qaa pue 245 GR
185. E definition pull down the Customize menu and release on the PSDC DATABASE definition menu item DECwindows Motif displays the Set Database dialog box This box lets you redirect editing and review of dump reports to an alternate database area You can also redirect performance analysis to an alternate area Specify PSDC Database Directory Equivalence name s USERSEB amp DECPS DATABASE Logical Name Table O Process table lt gt Job table Q Group table System table Cj Reset Dvassign cancel Helo The translation of the lowest level definition is displayed along with a toggle setting indicating the logical name table in which it is defined Chapter 8 Use the DECwindows Motif Interface 305 How You Customize To remove a definition click Deassign The dialog box is removed and the definition at the next highest level goes into effect possibly reloading a schedule and parameter file To create a new definition click a toggle to specify which logical name table the definition should be placed in and type in the new definition Click OK when done The dialog box is removed and any new schedule and parameter files are loaded There must be at least one definition of PSDC DATABASE defined for the interface to run Customize Parameters The DECwindows Motif Interface lets you view and change the Performance Manager parameters file The Performance Manager Parameters file contains workload definitions fa
186. F71 900000 Same as previous TD I45 RD33 500000 Same as previous TD I46 ESE20 2000000 Same as previous TD 147 TU56 750000 Same as previous TD 148 RZ22 1000000 Same as previous TD I49 RZ23 1000000 Same as previous TD I50 RZ24 1200000 Same as previous TD 151 RZ55 1000000 Same as previous TD 152 RRD40S 120000 Same as previous TD 153 RRD40 120000 Same as previous TD I54 GENERIC DK 750000 Same as previous TD I55 RX23 400000 Same as previous TD I56 RF31 2100000 Same as previous TD 157 RF72 1200000 Same as previous TD I58 RAM DISK 5000000 Same as previous TD I59 RZ25 1800000 Same as previous TD I60 RZ56 1400000 Same as previous Chapter 5 Customize the Knowledge Base 145 Components of Rules 146 Performance Manager Administrator Guide Name TD I61 RZ57 TD I62 RX23S TD I63 RX33S TD RA92 TD I65 SSTRIPE TD I66 RZ23L TD I67 RX26 TD I68 RZ57I TD I70 RZ58 TD I71 SCSI MO TD I72 RRD42 TD 173 CD LOADER 1 TD 174 ESE25 TD 175 RFH31 TD 176 RFH72 TD 177 RF73 TD 178 RFH73 TD 179 RA72 TD I80 RA71 TD I81 RF35 Default Value 1700000 400000 400000 1900000 1200000 1200000 500000 1500000 3100000 750000 120000 120000 2000000 2100000 1600000 2100000 2100000 1400000 1400000 2600000 Description Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as
187. FAULTS Free page faults FREEFAULTS Modified page faults MFYFAULTS Global page faults GVALID Hard page faults PREADIO System page faults SYSFAULTS TYPE FILECACHE File Cache Usage Plots the file operation attempt rate to the file system caches categorized by Hits FILE CACHE HIT Misses FILE CACHE MISS TYPE JOBS Number of Jobs Plots the number of processes categorized by Interactive INTERACTIVE PROCESSES Batch BATCH PROCESSES Network NETWORK PROCESSES Detached DETACHED PROCESSES TYPE LOCKS Distributed Locking Plots the number of distributed lock operations per second categorized by Incoming enqueues and converts INCOMING LOCKING Outgoing enqueues and converts OUTGOING LOCKING Local enqueues and converts LOCAL LOCKING TYPE MEMORY_UTILIZATION Memory Utilization Plots physical memory usage categorized by Percentage allocated to the free list FREELIM Percentage in the modified page list MODIFIED Percentage allocated to user processes USERWS Percentage allocated to system processes SYSTEMWS Percentage allocated to OpenVMS VMSALLOC 218 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE PROCESSES Number of Processes by State Plots the number of processes categorized as Inswapped active INPROCACT Inswapped inactive INPROCINACT Outswapped inactive OUTPROCINACT Outswapped active OUTPROCACT TYPE RESPO
188. G Average number of processes in free page wait state FREECNT Free list page count FREEFAULTS Number of free list page faults per second FREELIM Percent of physical memory allocated to the free list by the SYSGEN parameter FREELIM FREELIST Percent of physical memory on the FREELIST excluding the number of pages for FREELIM GVALID Global page faults per second HIB Average number of processes in hibernate wait state HIBO Average number of processes in hibernate outswapped wait state IDLE Percent CPU time that is idle time IMAGE ACTIVATIONS INCOMING LOCKING INPROCACT 212 Performance Manager Administrator Guide Number of image activations per second Number of incoming ENQs or Lock Conversion CVTs from remote nodes per second Number of active inswapped processes Channel INPROCINACT ISWPCNT INTERACTIVE PROCESSES INTERRUPT INT COMQ IRP CNT IRP MAX KERNEL LAT TERMIO LEF LEFO LG RESPONSE LOCAL LOCKING LOCK CNT LOGNAM LRP CNT LRP MAX MBREADS MBWRITES MED RESPONSE ADVISE PERFORMANCE GRAPH Description Number of inactive inswapped processes Inswaps per second Number of interactive processes Percent CPU time spent on the interrupt stack Number of computable interactive processes Count of the IRPs in use Length of the IRP list Percent CPU time charged to kernel mode time Number of LAT terminal I O operations per second Average number of processes in local
189. HEPAR F ACP HDRCACHE The value of the SYSGEN parameter ACP HDRCACHE which sets the number of pages blocks for caching file header blocks for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN ACP MAPCACHEPAR F ACP MAPCACHE The value of the SYSGEN parameter ACP MAPCACHE which sets the number of pages blocks for caching index file bit map blocks for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Appendix C Performance Manager Data Cells 503 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN ACP QUOCACHEPAR F ACP QUOCACHE The value of the SYSGEN parameter ACP QUOCACHE which sets the number of quota file entries cached for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN ACP WORKSETPAR F ACP WORKSET The value of the SYSGEN parameter ACP WORKSET which sets the default size pagelets of a working set for an ACP for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC D
190. HRUPUT ON ANY CHANN EL to 1 for a given Performance Manager two minute record Number of bytes per second through any UNIBUS port required to set the data cell EXCESS THRUPUT ON ANY CHANN EL to 1 for a given Performance Manager two minute record Name TD KDA CHANNEL IO TD KDB CHANNEL IO TD MASSBUS CHANNEL IO TD GENERIC DISK IO TD GENERIC DISK THRUPU T TD T1 RKO6 TD T2 RKO7 TD T3 RP04 TD T4 RPO05 TD T5 RP06 TD T6 RM03 TD T7 RP07 TD T8 RPO7HT TD T13 RM80 Default Value 800000 900000 1700000 30 1000000 20 20 26 26 26 26 32 32 30 Components of Rules Description Number of bytes per second through any KDAB5O port required to set the data cell EXCESS THRUPUT ON ANY CHANN EL to 1 for a given Performance Manager two minute record Number of bytes per second through any KDB port required to set the data cell EXCESS THRUPUT ON ANY CHANN EL to 1 for a given Performance Manager two minute record Number of bytes per second through any MASSBUS port required to set the data cell EXCESS THRUPUT ON ANY CHANN EL to 1 for a given Performance Manager two minute record Default I O operation per second threshold for disks that are not covered in the translation table TD Tn x thresholds Default number of bytes per second threshold for disks that are not covered in the translation table TD In x thresholds I O per second threshold for disk type Data cell DISK IO RATE
191. IBILITY mode for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU COMPAT TALLY Derived This contains the sum of the values representing the percent of time in COMPATIBILITY mode for the current CPU subrecords which were selected by the most recent CPU SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU 424 Performance Manager Administrator Guide Performance Manager Data Cells CPU EXECCPU F EXEC This contains the value representing the percent of time spent in EXEC mode for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU EXEC TALLY Derived This contains the sum of the values representing the percent of time in EXEC mode for the current CPU subrecords which were selected by the most recent CPU SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU CPU_IDLECPU_F_NULL This contains the value representing the percent of time the CPU was idle for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU_IDLE_TALLY Derived This contains the sum of the values representing the percent of time the CPU was idle for the current CPU subrecords which were selected by the most recent CPU_SCAN routine operation Data Type TALLY Domains COM
192. IGATE command m Aterminal with DEC CRT characteristics such as a VT100 is needed for all other displays Any number of users with these resources can simultaneously run the Performance Manager Real time character cell displays Start the Character Cell Displays Character cell displays can be invoked for either a single node or multiple nodes of a cluster system To begin collecting data for all nodes in a cluster system and to display system metrics in realtime mode 1 Enter the following command ADVISE PERFORMANCE DISPLAY CHARACTER CELL PA displays a multi node screen 2 To view previously recorded data in playback mode use the BEGINNING qualifier For more information on how to use the ADVISE PERFORMANCE DISPLAY command see the chapter Performance Manager Commands see page 175 Control the Displays Once you have started a character cell display you can control the display and its characteristics with commands The following table shows the Performance Manager character cell commands available at the PSRT prompt These commands control which displays and metrics are shown on the terminal The keypad keys perform other functions such as selecting users nodes and metrics for display and controlling the playback of data Command Function CPU Provides a multi node display of CPU utilization 374 Performance Manager Administrator Guide Command DISKS DISPLAY x FREEZE STEP IMAGENAME name
193. IL LOC PRO TAP PAGEFILE PAGE READ RATEMET F PREADS Average number of pages per second read from the page files for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PAGEFILE PAGE WRITE RATEMET F PWRITES Average number of pages per second written to the page files for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PAGEFILE_READ_IO_RATEMET_F_PREADIO Average number of reads per second from the page files for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PAGEFILE_UTILIZATION Derived This contains the ratio of used not free to total pages in all paging files for the current interval for the local node Data Type NUMERIC 474 Performance Manager Administrator Guide Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP PAGEFILE_WRITE_IO_RATEMET_F_PWRITIO Average number of writes per second to the page files for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PAGES_ON_FREELISTMET_F_FREECNT Average number pages on the free page list for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PAGES_ON_MODIFIEDLISTMET_F_MFYCNT Average number pages on the modified page list for the local node for the
194. INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOTAL_FAULT_RATEMET_F_FAULTS This contains a number representing the total page faults per second for the local node for the current interval record Data Type NUMERIC Appendix C Performance Manager Data Cells 525 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP TOTAL_OF_WS_SIZES Derived This contains the value representing the total number of pages PPGCNT GPGCNT that all processes are using on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TRANSITION_FAULT_RATEMET_F_TRANSFLTS Average number of global page faults per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TROLLER_IS_ON Derived The value of the SYSGEN parameter MMG_CTLFLAGS has bit 1 set for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP USERMET_F_USER Average percentage of CPU time spent in User mode for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP USER_NAMEPRO_A_USERNAME This contains a string indicating the user name for which the current process subrecord pertains on the local node for the cu
195. IO IMAGES graph type e 191 TOP DISKIO USERS graph type e 191 TOP DISKIO WORKLOADS graph type e 191 TOP EXEC PROCESSOR graph type e 191 TOP FAULTING IMAGES graph type e 191 TOP FAULTING USERS graph type e 191 TOP FAULTING WORKLOADS graph type e 191 TOP FREEBLK DISKS graph type e 191 TOP FREEBLK VOLUMES graph type e 191 TOP HARDFAULTING IMAGES graph type e 191 TOP HARDFAULTING USERS graph type e 191 TOP HARDFAULTING WORKLOADS graph type e 191 TOP HSC DISK IO graph type e 191 TOP HSC DISK THRUPUT graph type e 191 Index 589 TOP HSC IO graph type e 191 TOP HSC TAPE IO graph type e 191 TOP HSC TAPE THRPUT graph type e 191 TOP HSC THRUPUT graph type e 191 TOP IMAGE ACTIVATIONS graph type e 191 TOP IMAGE VOLUME IO graph type e 191 TOP INTERRUPT PROCESSOR graph type e 191 TOP IO DISKS graph type e 191 TOP IO FILES graph type e 191 TOP IO RULE OCC graph type e 191 TOP IO VOLUMES graph type e 191 TOP KB MAP graph type e 191 TOP KB RC graph type e 191 TOP KB S graph type e 191 TOP KERNEL PROCESSOR graph type e 191 TOP MEMORY RULE OCC graph type e 191 TOP MGS R graph type e 191 TOP MGS S graph type e 191 TOP MP SYNCH PROCESSOR graph type e 191 TOP MSCPIO FILES graph type e 191 TOP PAGING DISKS graph type e 191 TOP PAGING FILES graph type e 191 TOP PAGING VOLUMES graph type e 191 TOP POOL RULE OCC graph type e 191 TOP PRCT FREE DISKS graph type e 191 TOP PRCT FREE VOLUMES graph type e 191 TOP PRCT USE
196. IO graph e 191 HSC Tape IO graph e 191 Thruput graph e 191 HSC Thruput graph e 191 Image Activations Users graph e 191 Activations Workload graph e 191 I O Operations e 191 Images Activated graph e 191 Volume IO graph e 191 Interrupt Stack Processor e 191 IO Rule Occurrences graph e 191 KB Sent Rate graph e 191 Kernel Mode Processor e 191 Kilobyte Mapped Rate graph e 191 Received Rate graph e 191 Memory Rule Occurrences graph e 191 Messages Received graph e 191 Sent graph e 191 MP Synch Mode Processor e 191 MSCP I O Operations files graph e 191 Operations Files graph e 191 Operations Disk Device graph e 191 Volume graph e 191 Percent Freeblks Disk Device graph e 191 Volume graph e 191 Percent Usedblks Disk Device graph e 191 Volume graph e 191 PG amp SWP Operations Disk Device e 191 Disk Volume graph e 191 Files e 191 Pool Rule Occurrences graph e 191 Queue Disk Device graph e 191 Disk Volume graph e 191 HSC Channel e 191 Read Operations Disk Device e 191 Disk Volume e 191 Files e 191 Resource Rule Occurrences graph e 191 Response Time Disk Device graph e 191 File graph e 191 Images graph e 191 Users graph e 191 Workloads e 191 Rule Occurrences graph e 191 Service Time Disk Volume graph e 191 Split Operations Disk Device graph e 191 Disk Volume graph e 191 Files graph e 191 Supervisor Mode Processor e 191 system use graphs e 294 pie charts e 294 Terminal Input Images grap
197. ISE PERFORMANCE COMPILE command The default file type is KB If the NORULES qualifier is specified no augmentation of the factory rules occur See also the Chapter Customize the Knowledge Base see page 135 Chapter 6 Performance Manager Commands 197 ADVISE PERFORMANCE GRAPH SCHEDULE day hour range NOday Specifies that a subset of Performance Manager data is to be used or not used if keyword negation is specified to generate graphs By default the Performance Manager selects all data between the BEGINNING time and the ENDING time or as specified with the DATES qualifier Where day SUNDAY MONDAY TUESDAY WEDNESDAY THURSDAY FRIDAY SATURDAY EVERYDAY WEEKDAYS or WEEKENDS hour Specified as m n where m and n are numbers from 0 to 24 range and m is less than n You can specify more than one hour range for a given day Hour range is mutually exclusive with the NO option If you omit a day keyword the data for that day is selected Data selection for individual days of the week can be inhibited by negating the keyword for example NOSUNDAY or for all of the days of the week by specifying the NOEVERYDAY keyword The values NO WEEKDAYS and NO WEEKENDS similarly can be used to enable or disable data selection for weekdays and weekends You must specify an hour range for any non negated day keyword Do not include an hour range if you are specifying a negated day keyword such as NOMONDAY
198. K FILE LOCAL SUMMARY PROCESS TAPE Process PROCESS Tape TAPE Note The CLUSTER domain references a sub record type that exists only in memory It does not exist in the Performance Manager data records If you write a rule in the LOCAL domain the data cells available to the rule conditions and rule evidence include all values from the metrics and parameters records for a single node The Performance Manager evaluates the rule once for each interval record 2 minutes for daily data user defined for history data for each node that is processed For example an hour of daily data at the default interval contains 30 records per node for the LOCAL domain If you write a rule in the PROCESS domain data cells available to the rule include all values from the process sub record Metric and parameter sub record data cells are also available from the PROCESS domain because they are part of the current interval record containing the given process sub record for the current node The Performance Manager passes each of the process sub records through all rules in the PROCESS domain for each interval record and for each node The number of process sub records that the Performance Manager examines depends on the number of processes logged in the time range system activity and for history files the workload classification scheme You must reference data in a process sub record from the PROCESS domain or from the LOCAL COMMUNICATION CONFIGUR
199. K SERVICE TIME TALLY Derived This contains the sum of the values representing the average number of milliseconds between the I O events START IO and END IO for all I Os for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK SPLIT IO RATEDEV F SPLIT This contains a value representing the average number of split I O requests per second to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK SPLIT IO TALLY Derived This contains the sum of the values representing the average split I O rate for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_SWAPPING_IO_RATEDEV_F_SWPOP This contains a value representing the average number of swapping I O requests per second to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK Appendix C Performance Manager Data Cells 443 Performance Manager Data Cells DISK SWAPPING IO TALLY Derived This contains the sum of the values representing the average swapper I O rate for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TA
200. LAIN RULES MYRULES This command produces an Analysis Report for the node DEMAND This report is for the current day using the beginning time of midnight 00 00 and the current time of day as the ending time In addition to the Performance Manager factory rules the Performance Manager use an auxiliary knowledge base The NOEXPLAIN qualifier indicates that the report contains only conclusions and recommendations omitting the rule conditions and the evidence Because the OUTPUT qualifier is not specified the report is displayed on the terminal ADVISE PERFORMANCE REPORT BRIEF ANALYSIS BEGIN 30 JUN 1996 10 00 END 30 JUN 1996 14 00 _ OUTPUT ZERO IN This command produces a Brief Analysis Report for the time period between 10 00 a m and 2 00 p m on June 30 1996 The OUTPUT qualifier directs the output to a file called ZERO IN RPT The Brief Analysis Report contains rule identifiers the percentage of time for which there were instances of rule occurrences during the reporting period the number of Performance Manager data records two minute records supporting the rule occurrence and a brief no more than one line synopsis of the problem statement A cluster wide synopsis follows the synopsis for each node 236 Performance Manager Administrator Guide ADVISE PERFORMANCE REPORT As you become more familiar with analysis reports the brief report may be sufficient on a daily basis ADVISE PERFORMANCE REPORT ANALYSIS PERFO
201. LE OPERATION TALLY Derived This contains the sum of the values indicating the number of I Os per second transferred to and from all hot files indicated by FILE DEVICE FILE DIRECTORY and FILE NAME for the current hot file subrecords which were selected by the most recent FILE SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FILE_PAGING_IO_RATEFIL_F_PAGOP This contains a value indicating the number of paging I O operations per second issued to the file indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME for the current interval Data Type NUMERIC Domains FIL Appendix C Performance Manager Data Cells 449 Performance Manager Data Cells FILE PAGING IO TALLY Derived This contains the sum of the values indicating the number of Paging I Os per second transferred to and from all hot files indicated by FILE DEVICE FILE DIRECTORY and FILE NAME for the current hot file subrecords which were selected by the most recent FILE SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FILE_READ_RATEFIL_F_RDCNT This contains a value indicating the number of read I O operations per second issued to the file indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME for the current interval Data Type NUMERIC Domains FIL FILE_READ_TALLY Derived This contains the sum of the values indicating the number of Read I Os
202. LOCKOUT is true 1 0 this cell will point to the most computable process subrecord which has a lower priority than the high priority process This process is probably unable to utilize the CPU because of a priority lockout Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO 522 Performance Manager Administrator Guide Performance Manager Data Cells TOP CPU PROC BPRI Derived This contains a value representing the base priority of the process subrecord with the highest CPU utilization on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TOP_CPU_PROC_CPU Derived This contains a value representing the percent of CPU utilization of the process with the highest CPU utilization on the local node for the current interval TOP CPU PROC X indicates this process Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TOP_CPU_PROC_X Derived This contains an index pointing to a process subrecord whose process has the highest CPU utilization on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP CW SCS NODE X Derived This contains an index pointing to the configuration subrecord for the remote SCS node for which the local node suffered the highest rate of credit waits for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target
203. LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA LOCK 23 2 12 6 0 0 19 8 4 6 0 0 24 0 12 8 0 0 0 1 0 0 0 1 9 5 0 0 0 0 0 0 44 0 0 0 0 0 0 0 10308 0 4835 0 Example 23 2 12 6 0 0 19 8 4 6 0 0 24 0 12 8 0 0 0 1 0 0 0 1 Data item eng local eng in enq out cvteng local cvtenq in cvtenq out deq local deq in deq out blkast local blkast in blkast out Description enq ops sec by Icl node for Icl Icks eng ops sec req by rem node for Icl Icks enq ops sec req by Icl node for rem Icks conv ops sec by Icl node for Icl Icks conv ops sec by rem nodes for Icl Icks conv ops sec by Icl node for remote Icks deq ops sec by Icl node for Icl Icks deq ops sec by remote nodes for Icl Icks deq ops sec by Icl node for remote Icks blking ast sec by Icl node for Icl Icks blking ast sec by rem nodes for Icl Icks blking ast sec by Icl node for rem Icks Position Length 91 7 100 7 109 7 118 7 127 7 136 7 145 7 154 7 163 7 172 7 181 7 190 7 Appendix E Output Format for ASCII CSV Data 551 Lock Statistics Data Record Example Data item 9 5 dirfunc in 0 0 dirfunc out 0 0 dickmsg in 0 0 dickmsg out Description Icks sec for directory ops by remote node for Icl directories Icks sec for directory ops by Icl node for Icl directories deadlock msg sec received from
204. MB1 Click the buttons at the bottom of the window to perform the following actions Clear Removes all the entries from display and restores all default values Create Adds the collection definition you have created to the list The software displays an error message if it finds an error in your entries Chapter 8 Use the DECwindows Motif Interface 297 How You Customize Modify Updates the collection definition with your changes Delete Removes the collection definition from the list and restores all default values Reset Returns all the parameters on display to their original values Change the Collection Definition Parameters To change the collection definitions enter any of the following parameters Collection Definition List The box in the upper left hand corner contains the names of the current collection definitions When you click an entry the parameters for that collection definition appear Collection name The Collection name is a text field that shows the name of the definition currently displayed When creating a new definition use this field to enter the name Names can be up to 20 characters in length Default working set Enter the working set quota This value is a decimal number greater than 1024 that sets both the working set quota and the working set extent if the value is higher than the default values By default the working set quota is 2048 and the working set extent is 20K You can override thi
205. Metric instance data specify e 369 Minimum data value bar graph setting of e 347 strip chart setting of e 348 Modify MEMutl e 76 pages e 75 Modify panel instruments e 346 Most Resident Images graph e 191 Users graph e 191 Workloads graph e 191 Motif Interface customize parameters e 306 history file descriptors e 316 set parameters e 319 workload definitions e 308 workload family definitions e 313 Multi I O was busy e 79 N Node definition list e 298 maximum e 94 minimum e 94 Nonpaged memory field in system configuration report e 72 Number of Jobs graph e 191 Processes graph e 191 time units strip chart setting of e 348 Numeric operators e 153 O Open files e 82 P Page faults e 79 faults panel e 330 file allocation panel e 334 Paged memory in system configuration report e 72 Pages read e 79 writen e 79 Panel hierarchy e 324 instrument edit menu functions of e 345 instrument metrics assign e 359 Index 583 instruments background specifying of e 367 close e 371 copy e 342 358 create e 342 357 delete e 342 358 edit e 345 filter metrics select e 359 launch e 364 menu removal of e 370 modify e 346 node specify e 369 options set e 366 remove e 342 rename e 342 save e 371 set auto displays of e 342 status set e 367 title specify e 368 manager access e 341 Parameter Editor start e 306 settings auto augment e 319 menu e 319 version limit e 319 Parameters file change e 306 cu
206. N 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP DISK ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA DISK 1 DUA30 0 0 0 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 44 2 2376153 1324938 0 0 Example Data item Description Position Length 1 DUA30 disk name device name 92 20 0 0 work avail of time work 115 7 was available for disk 0 0 total work total 4 of work 124 7 request Appendix E Output Format for ASCII CSV Data 553 Disk Statistics Data Record Example Data item Description Position Length 0 0 remote work of work 133 7 requests from remote systems 0 0 disk paging Yo of total work 142 7 marked as pagio 1 0 rem paging Yo of total work 151 7 marked as pagio from remote systems 0 0 disk_swping of total work 160 7 marked as swapio 1 0 rem_swping of total work 169 7 marked as swapio from remote systems 1 0 server of server s 178 7 work charged to this disk 0 0 disk_rate I O rate second 187 7 0 0 service_time service time in ms 196 7 0 0 response time response time in 205 7 ms 0 0 que_length average queue 214 7 length 44 2 space used average 9o space 223 7 used 2376153 max blocks average 232 7 maximum space for use 1324938 free blocks average free 241 7 space 0 0 read cdrps read operations 250 7 for this disk 554 Performance Manager Administrator Guide Serv
207. N RWIMG Derived This contains a value representing the count of images seen by the data collector in an image activation lock resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWLCK Derived This contains a value representing the count of images seen by the data collector in a lock ID resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWMBX Derived This contains a value representing the count of images seen by the data collector in a mailbox resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWMPB Derived This contains a value representing the count of images seen by the data collector in a modified page list busy resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 470 Performance Manager Administrator Guide Performance Manager Data Cells NO IMAGES SEEN IN RWMPE Derived This contains a value representing the count of images seen by the data collector in a modified page list empty resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWNPG Derived This contains a
208. NAME IMAGENAME UIC GROUP PROCESS NAME WORKLOAD NAME ACCOUNT NAME Description Provide process records subgrouped by the process mode Interactive Batch Network or Detached Provide process records subgrouped by the process s User name The fields UIC and ACCOUNT are also enabled when this key is specified Provide process records subgrouped by the process s Image Name The field IMAGE DIRECTORY is also enabled when this key is used in combination with PID Provide process records subgrouped by the process s UIC Group Provide process records subgrouped by the process name Provide process records subgrouped by the workload name You must specify CLASSIFY BY to indicate the workload family that you intend to use Provide process records subgrouped by the process s account name PID Provide process records subgrouped by the process s EPID The fields USERNAME PROCESS NAME UIC MODE and ACCOUNT are also enabled when this key is specified If both PID and IMAGENAME are used for the primary and secondary keys the WORKLOAD NAME and IMAGE fields are also enabled that is all fields are enabled 188 Performance Manager Administrator Guide ADVISE PERFORMANCE EXPORT If the value for secondary key is the same as for the primary key no secondary level breakout occurs This also happens if the primary key is specified and no secondary key is given The following example demonstrates the u
209. NDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO HSC_IO_RATE Derived This contains a value representing the disk operations rate of the current HSC controller Data Type NUMERIC Domains CLU HSC_NODE_NAME Derived This contains a text string representing the node name of the current HSC controller Data Type STRING Domains CLU 454 Performance Manager Administrator Guide Performance Manager Data Cells HSC THRUPUT RATE Derived This contains a value representing the disk throughput rate in bytes per second of the current HSC controller Data Type NUMERIC Domains CLU HSC TYPE HSC40 Derived This contains a Boolean value 1 if the current HSC controller is an HSC40 Data Type NUMERIC Domains CLU HSC TYPE HSC50 Derived This contains a Boolean value 1 if the current HSC controller is an HSC50 Data Type NUMERIC Domains CLU HSC TYPE HSC60 Derived This contains a Boolean value 1 if the current HSC controller is an HSC60 Data Type NUMERIC Domains CLU HSC TYPE HSC65 Derived This contains a Boolean value 1 if the current HSC controller is an HSC65 Data Type NUMERIC Domains CLU HSC TYPE HSC70 Derived This contains a Boolean value 1 if the current HSC controller is an HSC70 Data Type NUMERIC Domains CLU HSC TYPE HSC90 Derived This contains a Boolean value 1 if the current HSC controller is an HSC90 Data Type NUMERIC Appendix C Performance
210. NSE TIME Terminal Response Time Plots the terminal response time for interactive processes categorized as Large transactions LG RESPONSE Medium transactions MED RESPONSE Small transactions SM RESPONSE TYPE TERMINALS Terminal I O Plots the number of terminal operations per second categorized by the type of terminal used TX TX TERMIO TT IT TERMIO RT RT TERMIO LT LAT TERMIO NV NV TERMIO TW TW TERMIO TYPE TOP BDT W Top BDT Wait Rate Plots five remote nodes with the highest rate of BDT waits plus Other resulting when a local node issues an I O but the connection had to wait for a buffer descriptor The metric graphed is BDT W TYPE TOP BLKS R Top Bik Transfers Requested Plots the top five nodes with the highest block transfer requests plus Other from the remote system to the local system The metric graphed is BLKS R TYPE TOP BLKS S Top Bik Transfers Sent Plots the top five nodes with the highest block transfers sent plus Other from the local system to the remote system The metric graphed is BLKS S TYPE TOP BUFIO IMAGES Top Buffered I O Images Plots the top five plus Other Images creators of buffered I O by image names The metric graphed is BUFIO Chapter 6 Performance Manager Commands 219 ADVISE PERFORMANCE GRAPH TYPE TOP BUFIO USERS Top Buffered I O Users Plots the top five plus Other Users creators of buffered I O by user nam
211. O TAP LOGICAL_NAME_TRANSLATION_RATEMET_F_LOGNAM Average logical name translation rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LRPS_IN_LISTMET_F_LRP_MAX This contains the total number of LRPs for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LRPS_IN_USEMET_F_LRP_CNT This contains the number of LRPs in use for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 465 Performance Manager Data Cells LRP EXPANSION COUNT Derived A count of the number of times that the number of large request packets needed to be increased for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY LRP MAXLEN Derived The maximum size of the LRP list for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY MAILBOX READ RATEMET F MBREADS Average mailbox read rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP MAILBOX_WRITE_RATEMET_F_MBWRITES Average mailbox write rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP MASTER_PIDPRO_L_MPID This contains a hexadecimal representation of the master PI
212. OC PRO TAP HEAD IN SWAP RATEMET F HISWPCNT Average process header inswap rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP HEAD OUT SWAP RATEMET F HOSWPCNT Average process header outswap rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP HIGHEST_IO_RATE_DISK_X Derived This contains an index pointing to the disk subrecord that has the highest I O operations rate on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK Appendix C Performance Manager Data Cells 453 Performance Manager Data Cells HIGHEST QUEUE DISK X Derived This contains an index pointing to the disk subrecord that has the highest queue on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK HIGHEST_SPLITIO_RATE_DISK_X Derived This contains an index pointing to the disk subrecord that has the highest split I O operations rate on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK HIGH_IMG_ACTIVATIONS_PID_X Derived This contains an index pointing to the process subrecord whose PID has the highest number of image activations on the local node for the current interval Data Type I
213. OC PRO TAP Target Domains PRO 528 Performance Manager Administrator Guide Performance Manager Data Cells WORKING SET EXTENTPRO F WSEXTENT This contains a value representing the current working set extent of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO WORKING SET EXTENT TALLY Derived This contains the sum of the values representing the number of pages allocated as working set extents for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO WORKING SET FAULT IO RATEPRO F PGFLTIO This contains a value representing the hard page fault rate per CPU second for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO WORKING SET FAULT IO TALLY Derived This contains the sum of the values representing the hard page fault rate per CPU second for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO WORKING SET FAULT RATEPRO F PAGEFLTS This contains a value representing the soft page fault rate per CPU second for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO Appendix C Performance
214. ON RATE TALLY 160 Performance Manager Administrator Guide Scan Routines CONFIGURATION SCAN CPU SCAN DISK SCAN FILE SCAN Data Cell Types and Use Tally Data Cells Updated by Scan Routines DATAGRAMS SEND TALLY DATAGRAMS RECEIVED TALLY DATAGRAMS DISCARDED TALLY SEQUENCED MESSAGES SENT TALLY SEQUENCED MESSAGES RECD TALLY BLOCK SEND DATAS INIT TALLY KB SENT VIA SEND DATAS TALLY BLOCK REQUEST DATAS INIT TALLY KB RECVD VIA REQST DATAS TALLY KB MAPPED TALLY SEND CREDIT QUEUE TALLY BUFFER DESC QUEUE TALLY CPU KERNEL TALLY CPU EXEC TALLY CPU SUPER TALLY CPU USER TALLY DISK SERVICE TIME TALLY DISK QUEUE LENGTH TALLY DISK IO RATE TALLY DISK THRUPUT TALLY DISK PAGING IO RATE TALLY DISK PAGING THRUPUT TALLY DISK SWAPPING IO TALLY DISK SWAPPING THRUPUT TALLY DISK BUSY PERCENT TALLY DISK ERROR COUNT TALLY DISK READ IO RATE TALLY DISK FREE PAGES TALLY DISK MSCP IO RATE TALLY DISK MSCP PAGING IO TALLY DISK MSCP THRUPUT TALLY DISK SPLIT IO TALLY FILE THROUGHPUT TALLY FILE OPERATION TALLY FILE READ TALLY FILE SPLIT IO TALLY FILE PAGING IO TALLY FILE SWAPPING IO TALLY Chapter 5 Customize the Knowledge Base 161 Data Cell Types and Use Scan Routines PROCESS SCAN TAPE SCAN Index Specifier Data Cell Tally Data Cells Updated by Scan Routines PROCESS CPUTIME TALLY WORKING SET FAULT TALLY WORKING SET FAULT IO TALLY PROCESS DIRECT IO TALLY PROCESS BUFFERED IO TALLY GLOBAL PGS TALLY PRIVATE PGS
215. ORMANCE GRAPH BEGINNING 19 FEB 2006 13 00 _ ENDING 19 FEB 2006 17 00 TYPE CPU UTILIZATION NODE GALLO _ FORMAT POSTSCRIPT CHARACTERISTICS LINE OUTPUT CPU LINE PS 118 Performance Manager Administrator Guide Format Graphs and Pie Charts The LINE keyword is used with the NOSTACK qualifier to avoid occlusion CPU UTILIZATION 20 Wade GALLO Dae 12 2E2 139T 1300 17 00 AE Ptr tack 3 0 Percent of CPU n bi gt m gt 7h F gt Chapter 4 Generate Historical Graphs 119 Format Graphs and Pie Charts Y Units Pe II II D N D D ID IN D DNIDID DIDDDD rcent N B I I D D D DI D DID DID DID DNDDDI NDNDDDI NNDIDDDI N I I D ANSI Formatted Graph is the same CPU utilization graph as Tabular Formatted Graph in ANSI formatted output The default width of an ANSI graph is 132 characters This graph overrides the default width It was generated with the following command ADVISE PERFORMANCE GRAPH BEGINNING 19 FEB 2006 13 00 ENDING 19 FEB 2006 17 00 TYPE CPU UTILIZATION NODE GALLO OUTPUT CPU LIS _ FORMAT ANSI WIDTH 79 HEIGHT 25 CPU UTILIZATION Node GALLO Date 19 FEB 2006 13 00 17 00 LEGEND Other N Network B Batch I Interactive D Detached X Intstk MPsynch Metric Values are Stacked N NNNNNNNNNNNNNNNNNNN N NNNNNBINNNBNINNBNBNNNNB NNNNBIBIIIIIBIIIBBIBIIIIBI INIIIIIIIIIIIIIIIIIIIIIIII N IIIIIIIIIIIIIIIIIIIIIIIIII IN
216. P PROCESS NAME WORKLOAD NAME ACCOUNT NAME PID L NO CLUSTER L NO BY NODE The previous code lets you tailor the process statistics section of the Performance Evaluation report You can specify the focus of the report to obtain slightly different sorts of statistics that pertain to the focus area The grouping merging and sorting of the process data is controlled with the primary and secondary key settings To use a given primary or secondary key you must have previously specified the process key level with the SELECT command See the description of the PERFORMANCE EVALUATION option with SELECT Also you can specify whether a cluster wide report or a node by node presentation is desired By default the focus area is TRADITIONAL that being an image based report showing relative resource consumption Reports are provided by node unless otherwise specified Where FOCUS Types Provide TRADITIONAL An 80 column report showing process CPU memory and IO statistics on a per image activation basis or as a relative percentage The default primary and secondary keys are MODE and IMAGENAME The SUMMARY focus report is also provided with the TRADITIONAL flavor 252 Performance Manager Administrator Guide FOCUS Types SUMMARY GENERAL MEMORY RELATED IO RELATED CPU RELATED Primary Keys MODE USERNAME IMAGENAME UIC GROUP PROCESS NAME WORKLOAD NAME ACCOUNT NAME PID Secondary Keys MODE USERNAME
217. P Target Domains DSK DISK_SWAPPING_THRUPUT_RATEDEV_F_SWPIO This contains a value representing the average number of bytes per second for swapping I Os transferred to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_SWAPPING_THRUPUT_TALLY Derived This contains the sum of the values representing the average swapper throughput rate for the current disk subrecords which were selected by the most recent DISK_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_THRUPUT_RATEDEV_F_IOCNT This contains a value representing the average number of bytes per second transferred to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_THRUPUT_RATE_THRESHOLD Derived This contains a value representing the disk throughput rate threshold for the current disk This value is obtained from the threshold TD In xxxx where n is the disk type model number found in STARLET DCDEF for the current disk and xxxx is its type e g TD 121 RAB81 Data Type NUMERIC Domains CLU 444 Performance Manager Administrator Guide Performance Manager Data Cells DISK THRUPUT TALLY Derived This contains the sum of the values representing the average throughput rate for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type
218. PE TOP HSC TAPE THRUPUT Top HSC Tape Thruput Plots the top five plus Other HSCs in terms of tape thruput in Kilobytes per second The metric graphed is HSC TAPE THRUPUT TYPE TOP HSC THRUPUT Top HSC Thruput Plots the top five plus Other HSCs in terms of total thruput in Kilobytes per second The metric graphed is HSC THRUPUT TYPE TOP IMAGE ACTIVATIONS Top Images Activated Plots the top five plus Other images in terms of image activations per second The metric graphed is IMAGE ACTIVATIONS TYPE TOP IMAGE VOLUME IO Top I O Images and the Disk Volumes they access Plots the top five plus Other image and volume name pairs in terms of their I O rate The metric graphed is IMAGE VOLUME IO Chapter 6 Performance Manager Commands 223 ADVISE PERFORMANCE GRAPH TYPE TOP INTERRUPT PROCESSOR Top Interrupt Stack Processor Plots the five plus Other processors in terms of time spent on the interrupt stack as a percent of CPU time The metric graphed is P INTERRUPT TYPE TOP IO DISKS Top Operations Disk Device Plots the five plus Other Disks disk devices that incurred the highest I O rates The metric graphed is TOTIO TYPE TOP IO FILES Top IO Operations Files Plots the five plus Other files that incurred the highest I O rates The metric graphed is FILE TOTIO TYPE TOP IO RULE OCC Top IO Rule Occurrences Plots the five plus Other IO rule identifiers that fire
219. POSITE qualifier If you omit the NODE NAMES qualifier all the nodes in the schedule file associated with the specified collection definition CPD by default are used for the graph s If you specify only one node the parentheses can be omitted Do not use wildcard characters in the node name specifications OUTPUT filespec Creates an output file that contains the graphs The default file extension for a ReGIS graph is REG the file type for ANSI and TABULAR formatted graphs is RPT and the file extension for PostScript is PS When you generate multiple graphs with a single command line you can create a unique output file for each graph To do this omit the file name with the OUTPUT qualifier The Performance Manager generates a separate file for each graph created and uses the graph type keyword as the unique file name For example ADVISE PERFORMANCE GRAPH NODE SYSDEV END 1 TYPE MEM CPU U CPU MODE OUTPUT REG PSPA I CREAGRAPHOUT PSPA Graph created file MUMMS DKA300 CORREY SYSDEV CPU UTILIZATION REG 1 PSPA I CREAGRAPHOUT PSPA Graph created file MUMMS DKA300 CORREY SYSDEV MEMORY UTILIZATION REG 1 PSPA I CREAGRAPHOUT PSPA Graph created file MUMMS DKA300 CORREY SYSDEV CPU MODES REG 1 RULES file spec NORULES Loads information from the rules file to establish user defined hardware scaling factors The file spec must point to an auxiliary knowledge base which has previously been compiled with the ADV
220. PRIMARY KEY MODE USERNAME IMAGENAME UIC GROUP PROCESS NAME WORKLOAD NAME ACCOUNT NAME PID3 Key option MODE USERNAME IMAGENAME UIC GROUP PROCESS NAME WORKLOAD NAME Description Group process statistics by the process mode Interactive Batch Network or Detached Group process statistics by the process s User name The fields UIC and ACCOUNT are also enabled when this key is specified Group process statistics by the process s Image Name The field IMAGE DIRECTORY is also enabled when this key is used in combination with PID Group process statistics by the process s UIC Group Group process statistics by the process name Group process statistics by the workload name You must specify CLASSIFY BY to indicate the workload family that you intend to use ACCOUNT NAME Group process statistics by the process s account name Chapter 6 Performance Manager Commands 187 ADVISE PERFORMANCE EXPORT Key option PID Description Group process statistics by the process s EPID The fields USERNAME PROCESS NAME UIC MODE and ACCOUNT are also enabled when this key is specified If both PID and IMAGENAME are used for the primary and secondary keys the WORKLOAD NAME and IMAGE fields are also enabled i e all fields are enabled Secondary key options are SECONDARY KEY MODE USERNAME IMAGENAME UIC GROUP PROCESS NAME WORKLOAD NAME ACCOUNT NAME PID3 Secondary key option MODE USER
221. PROC YQUEM Cluster Summary Statistics with By Node Breakout Tabular cluster reports have two formats by cluster and by node The following example shows final statistics by cluster In both formats statistics are given for memory CPU disks and locks Chapter 3 Evaluate Performance in Detail 93 Tabular Report Sections Tabular Report CLUSTER PA Vx x Tuesday 26 JAN 2006 09 00 to 09 20 FRR A 2 2 2 2K ok EE KK E k E k OK ok k kk kk Final Statistics 2 ES 2 2 2 2 2 2 2 E 2 KE 2 ke 2K 2K k k k kkk pbe Data Analyzed from 26 JAN 2006 09 00 00 00 to 26 JAN 2006 09 20 00 FEC IOI I AOE Total Memory Balset Page Hard Soft Gvalid MEMutl Queue Count Faults Faults Faults Faults avg sec sec Node Average Node Minimum Node Maximum Cluster Total CLUSTER I 0 CPU System Task CPU CPU CPU IO CPU I 0 Multi CPU IO Direct Buffrd Busy CPU CPU Queue Idle Idle Only Only I O Busy I 0s 5 5 35 Node Average Node Minimum Node Maximum Cluster Total H Orig Out Lck Act Bound 55 sec Node Average Node Minimum Node Maximum Cluster Total User Command ADVISE PERF REPORT TABULAR BYCLUSTER BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 09 20 00 00 OUT BYCLU TAB SECTION SUMMARY Memory By Cluster Format In the By Cluster format for summary statistics the leftmost column of the tabular report contains the words Node Average Node Minimum Node Maximum and Cluster Total Each
222. PU DSK FIL LOC PRO TAP INTERRUPT_STACKMET_F_INTSTK Average percentage of CPU time on the Interrupt Stack for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IN_SWAP_RATEMET_F_ISWPCNT Average inswap rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IO_ONLYMET_F_SPMIOONLY Percentage of time that the CPU or all CPUs in a multiprocessing system were idle and at least one disk device was busy for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IRPS_IN_LISTMET_F_IRP_MAX This contains the total number of IRPs for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IRPS_IN_USEMET_F_IRP_CNT This contains the number of IRPs in use for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 459 Performance Manager Data Cells IRP EXPANSION COUNT Derived A count of the number of times that the number of intermediate request packets needed to be increased for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY IRP MAXLEN Derived The maximum size of the IRP list for the local node for all of the intervals Data Type NUMERIC Domains
223. Performance Manager Data Cells NODENAME Derived Name of the local node that the other data cells in the same domains refer Data Type STRING Domains COM CFG CPU DSK FIL LOC PRO TAP NODE_INDX Derived This contains a sequence value representing the current node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NONPRIMARY_IDLE Derived Average percentage of CPU time idle for the processors other than the primary processor in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO_IMAGES_SEEN_IN_RWAST Derived This contains a value representing the count of images seen by the data collector in an AST resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN IN RWBRK Derived This contains a value representing the count of images seen by the data collector in a breakthrough resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 469 Performance Manager Data Cells NO IMAGES SEEN IN RWCLU Derived This contains a value representing the count of images seen by the data collector in a cluster transition resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NO IMAGES SEEN I
224. Performance Manager process since any unneeded pages are not wasted Appendix D Estimate Virtual Memory Needs 537 Appendix E Output Format for ASCII CSV Data This appendix describes the format of the data file you create when you use the ADVISE PERFORMANCE EXPORT TYPE ASCII command The appendix shows examples of each data record followed by a table showing each example value data item and a description of each item The data is in CSV Comma Separated Variable format with each item appearing as a fixed length field within the record In an actual data file the version record appears first indicating the node processed data version and image Each subsequent data record appears on a line beginning with the record header In the examples where a single data record line is too long to show here the record appears on several lines The records are described in the following sections m Record Header m Version Data Record m Memory Statistics m CPU Statistics Data Record m Secondary CPU Statistics Data Record m Page Statistics Data Record I O Statistics Data Record m XQP Statistics Data Record m System Communications Services Data Record m Lock Statistics Data Record m Device Statistics Data Record m Disk Statistics Data Record Appendix E Output Format for ASCII CSV Data 539 Record Header m Service Statistics Data Record m Process Metric Statistics Data Record This section contains the following topics
225. Performance Manager records to select data for the CUSTOM graph If you specify TYPE CUSTOM USER METRICS CPUTIME SELECTION WILK the Performance Manager selects and graph all process records which have the username field WILK The CUSTOM graph type allows you to graph a selection of metrics for either the system or selected users images workloads disk devices volumes HSCs SCS nodes rule ids or channels You may graph up to six selections with a single metric or up to six metrics with a single selection The Performance Manager either prompts you in command mode for the data ADVISE PERFORMANCE or you can specify the desired metrics and selections in a single DCL command For example ADVISE PERFORMANCE GRAPH TYPE CUSTOM SYSTEM METRICS DZROFAULTS GVALID The SELECTION string must be chosen based on the metric class that you use If the metric class is USER METRICS then the selection strings are interpreted as user names If the metric class is IMAGE METRICS then the selection strings are interpreted as image names Ifthe metric class is WORKLOAD METRICS then the selection strings are interpreted as workload names Unless you use the CLASSIFY BY qualifier to characterize the process data into various workloads all workload data is grouped into the default workload called OTHER Ifthe metric class is DEVICE METRICS then the selection strings are interpreted as device names If the met
226. Queued Metric related to sequenced messages The number of times that a local node Sysap system application had to wait for sufficient credits on the target node to become available to complete a transfer The number of credits is controlled by the system parameter SCSRESPCNT 86 Performance Manager Administrator Guide Tabular Report Sections Send Data Metric related to block transfers The number of times per second that data was written to a remote node using block mode transfers that were initiated by the local node This field is zero for all other nodes in the list K Byte Sent The amount of information in Kbytes written to some remote node using block mode transfers that were initiated by the local node Used for block transfers primarily HSC MSCP and Connection Manager transfers This field is zero for all other nodes in the list A process on a remote node does a QIO read from a disk that is hosted locally The MSCP server on the local node has to read the data from the local disk then do an SCS Send Data Directive that initiates a block mode transfer to write that data to the remote node Reqst Data Metric related to block transfers The number of times per second that data was read from a remote node using block mode transfers that were initiated by the local node This field is zero for all other nodes in the list K Byte Reqd The amount of information in Kbytes read from a remote node using block mode
227. R VOLUMENAMES to graph all disk records with the volumename matching any of the specified strings Do not specify any trailing colon DEVICENAMES string Specify FILTER DEVICENAMES to graph all disk records with the devicename matching any of the specified strings Do not specify any trailing colon FORMAT ReGIS CHARACTERISTIC COLOR LINE PATTERN LX POINTS 1 ANSI HEIGHT m WIDTH n LINE TABULAR X POINTS I CSV X POINTS I POSTSCRIPT CHARACTERISTIC COLOR LINE PATTERN X POINTS I Where Is in the range of 2 to 480 and a best fit value is chosen by default m Is greater than or equal to 20 and less than or equal to 60 n Is greater than or equal to 40 and less than or equal to 132 The Performance Manager graphs ReGIS or ANSI graph by default depending on the device characteristics of the SYSSOUTPUT device ANSI and ReGIS formats are not available with pie charts You may override the default with the FORMAT qualifier A graph is one of four formats ANSI REGIS TABULAR or PostScript Chapter 6 Performance Manager Commands 195 ADVISE PERFORMANCE GRAPH Optionally you may specify whether ReGIS graphs use LINE PATTERN or COLOR COLOR is the default PATTERN is incompatible with COLOR Use the X POINTS keyword to specify the number of data points to plot across a ReGIS graph The valid range for X POINTS is 2 to 480 By default the Performance Manager
228. RMANCE EVALUATION _ BEGIN 10 00 END 14 00 _ OUTPUT ZERO IN This command produces both an Analysis Report and a Performance Evaluation Report for the nodes listed in the schedule file These reports are for the time period between 10 00 a m and 2 00 p m on the current day The OUTPUT qualifier directs the output to a file called ZERO IN RPT ADVISE PERFORMANCE REPORT HISTOGRAM PERFORMANCE EVALUATION _ PROCESS STATISTICS _ PRIMARY KEY USERNAME SECONDARY KEY IMAGENAME _ INCLUDE PROCESS END 8 IMAGE LOGINOUT OUTPUT CHECK BREAKIN This command produces a Performance Evaluation Report for the current day from midnight to 8 00 a m An additional histogram for the LOGINOUT image is generated The OUTPUT qualifier writes the Performance Evaluation Report to a file called CHECK_BREAKIN RPT Only the Process Statistics section of the Performance Report is produced showing the process activity of each user by image ADVISE PERFORMANCE REPORT TABULAR INTERVAL _ INTERVAL 600 BEGIN 10 00 END 11 00 _ NODE MYNODE SECTION SUMMARY STATISTICS This command produces the summary statistics section of the tabular report for the node MYNODE The section is repeated 6 times each summarizing 10 minutes of data from within the 1 hour reporting period Chapter 6 Performance Manager Commands 237 ADVISE PERFORMANCE SHOW VERSION ADVISE PERFORMANCE SHOW VERSION Use the ADVISE PERFORMANCE SHOW VERSION command to display the
229. RO PROCESS WAS IN HIBOPRO V SSS HIBO This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the outswapped hibernate scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN LEFPRO V SSS LEF This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the local event flag wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Appendix C Performance Manager Data Cells 489 Performance Manager Data Cells Domains PRO PROCESS WAS IN LEFOPRO V SSS LEFO This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the outswapped local event flag wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN MWAITPRO V SSS MWAIT This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the mutex wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN PF
230. S 202 Performance Manager Administrator Guide NO TOP HARDFAULTING WORKL OADS NO TOP HSC DISK THRUPUT NO TOP HSC TAPE IO NO TOP HSC THRUPUT NO TOP IMAGE ACTIVATIONS NO TOP INTERRUPT PROCESSO R NO TOP IOSIZE VOLUMES NO TOP IOSIZE USERS NO TOP IO DISKS NO TOP IO RULE OCC NO TOP KB MAP NO TOP KB S NO TOP MEMORY RULE OCC NO TOP MGS S NO TOP MSCPIO FILES NO TOP PAGING FILES NO TOP POOL RULE OCC NO TOP PRCT FREE VOLUMES NO TOP PRCT USED VOLUMES NO TOP QUEUE VOLUMES NO TOP READ FILES NO TOP RESIDENT IMAGES NO TOP RESIDENT WORKLOAD S NO TOP RESPONSE TIME DISK S NO TOP RESPONSE TIME IMAG ES NO TOP RESPONSE TIME VOLU MES NO TOP RULE OCCURRENCES NO TOP SPLITIO FILES NO TOP SUPER PROCESSOR ADVISE PERFORMANCE GRAPH NO TOP TERMINAL INPUT IMAGES NO TOP TERMINAL INPUT USER S NO TOP TERMINAL INPUT WORKLOAD NO TOP TERMINAL THRUPUT S IMAGES NO TOP TERMINAL THRUPUT USERS NOJ TOP TERMINAL THRUPUT WORKLOADS NO TOP THRUPUT DISKS NO TOP THRUPUT FILES NO TOP THRUPUT IMAGES NO TOP THRUPUT USERS NO TOP THRUPUT VOLUMES NO TOP THRUPUT WORKLOADS NO TOP USER IMAGE ACTIVATIONS NO TOP USER PROCESSOR NO TOP USER VOLUME IO NO TOP VA IMAGES NO TOP VA USERS NO TOP VA WORKLOADS NO TOP WORKLOAD IMAGE ACTIVATI NO TOP WRITE DISKS ONS NO TOP WRITE FILES NO TOP WRITE VOLUMES NO TOP WSSIZE IMAGES NO TOP WSSIZE USERS NO TOP WSSIZE WORKLOADS The follo
231. S Top Workload Terminal Thruput Plots the top five plus Other Workloads workloads with the highest character per second terminal thruput The metric graphed is TERM THRUPUT TYPE TOP THRUPUT DISKS Top Throughput Disk Device Plots the five plus Other Disks disk devices that incurred the highest throughput rates The metric graphed is THRUPUT TYPE TOP THRUPUT FILES Top Throughput Files Plots the five plus Other files that incurred the highest throughput rates The metric graphed is FILE THRUPUT TYPE TOP THRUPUT IMAGES Top Throughput Images Plots the five plus Other images with the highest throughput rates The metric graphed is THRUPUT 228 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE TOP THRUPUT USERS Top Throughput Users Plots the five plus Other users with the highest throughput rates The metric graphed is THRUPUT TYPE TOP THRUPUT VOLUMES Top Throughput Disk Volume Plots the five plus Other disk volumes that incurred the highest throughput rates The metric graphed is THRUPUT TYPE TOP THRUPUT WORKLOADS Top Throughput Workloads Plots the five plus Other workloads with the highest throughput rates The metric graphed is THRUPUT TYPE TOP USER IMAGE ACTIVATIONS Top Image Activations Users Plots the top five plus Other users in terms of image activations per second The metric graphed is IMAGE ACTIVATIONS
232. S X POINTS number Chapter 4 Generate Historical Graphs 123 Format Graphs and Pie Charts Command mode PSPA gt SELECT GRAPH X POINTS number m DECwindows Specify the Additional Options option from the Select Data dialog box As the value of X POINTS increases more peaks and valleys appear on a graph As the value decreases the peaks and valleys are smoother because Performance Manager averages data points within the time frame requested The next three figures illustrate the relationship between the value of X POINTS and a time frame of 4 hours During this period the Performance Manager records statistics 120 times every 2 minutes The following graph X POINTS Default Value Graph uses 60 for the value of X POINT Therefore two data records are averaged to calculate the value of each point plotted This command generates the graph ADVISE PERFORMANCE GRAPH BEGINNING 19 FEB 2006 13 00 ENDING 19 FEB 2006 17 00 _ TYPE CPU UTILIZATION NODE GALLO OUTPUT SUPPLY XP 60 PS _ FORMAT POST CHARACTERISTIC PATTERN CPU UTILIZATION Vio Node GALLO Date 19 FEB 1997 13 00 17 00 48 Pix Each 5 0 Ino KO Percent of CPU 40 SAIYIVA GINDVLS 1300 13 15 13 30 13 45 14 00 14 15 14 30 14 45 15 00 15 15 15 30 15 45 16 00 16 15 16 30 16 45 Min 16 87 Max 100 00 Average 62 51 Batch Detached aue p Interactive ane Intsk MPeynch E Other Bes 124 Performance Manager Administrator Guide Forma
233. S to select all disk records with the volumename matching any of the specified strings Do not specify any trailing colon FILTER DEVICENAMES string Specify FILTER DEVICENAMES to select all disk records with the devicename matching any of the specified strings Do not specify any trailing colon HISTORY history descriptor nam Specifies the name of a history file descriptor to cause history files to be used instead of daily data By default no history selection is made NODE_NAMES nodename Specifies the list of node names on which to select data By default the Performance Manager uses all nodes RULES file Specifies a user compiled rules file to be used when data is selected for Analysis SCHEDULE dow m n Specifies that a weekly selection schedule is to be used when selecting data By default no schedule is used X_POINTS n Specifies the number of points to plot along the x axis for graphs X_POINTS also affects the width of ANSI formatted graphs The default value varies depending on the time period selected 244 Performance Manager Administrator Guide LOAD Format GRAPH Format Description Qualifiers LOAD The LOAD command allows you to load a selection of graph data that was previously saved to the specified file If you already have a period of time selected this command replaces the current selection LOAD file spec The GRAPH command graphs any group o
234. SE PERFORMANCE GRAPH The previous command produces one composite graph of archived data for all nodes in the cluster system ADVISE PERFORMANCE PSPA gt SELECT BEGIN 12 00 PSPA gt GRAPH TYPE PROMPT Please select either 1 a predefined graph or 2 a custom graph Choice In command mode GRAPH TYPE PROMPT displays available graph types and custom metrics as shown in the previous command Chapter 6 Performance Manager Commands 231 ADVISE PERFORMANCE PIE CHART ADVISE PERFORMANCE PIE CHART Use the ADVISE PERFORMANCE PIE CHART command to produce a pie chart instead of a graph The PIE CHART option has the same format as the ADVISE PERFORMANCE GRAPH command however the data is presented as a pie chart instead of as a graph Format ADVISE PERFORMANCE PIE CHART Description The Performance Manager can produce a multitude of predefined charts however only PostScript and DECwindows formats are supported The pie charts have the advantage of being able to display more than the top 5 values Qualifiers The following are qualifiers that are specific to pie charts For a complete description of the remaining ADVISE PERFORMANCE PIE CHART qualifiers refer to the ADVISE PERFORMANCE GRAPH see page 191 command FILTER keyword The FILTER qualifier allows you select a subset of the daily or history data for charting Process data and disk data can be filtered Hot file data is also filtered When you specify filtering by process
235. SKIO_PROCESS_X Derived This contains an index pointing to the process subrecord that has the highest disk I O operations rate on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO 524 Performance Manager Administrator Guide Performance Manager Data Cells TOP DSKIO PROC TOPDSK X Derived This contains an index pointing to the disk record with the highest operations rate for the process subrecord that has the highest disk I O operations rate on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK TOP_HF_IMAGE_X Derived This contains an index pointing to the process subrecord whose image has the highest hard page fault rate of all process subrecords on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP_HF_USER_X Derived This contains an index pointing to the process subrecord whose user has the highest hard page fault rate of all process subrecords on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP_QLEN_DISKS_PROCESS_X Derived This contains an index pointing to the process subrecord that has the highest I O operations rate to the disk with the highest queue length on the local node for the current interval Data Type
236. SS WS GTR QUOTA EXIST Derived This contains a Boolean value representing the presence of one or more processes where the working set is greater than its WSQUOTA on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESS WS GTR QUOTA PROC X Derived This contains an index pointing to the process subrecord for a unique username for processes where the working set is greater than its WSQUOTA on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROC_NOT_USING_WS_LOAN_X Derived This contains an index pointing to a process subrecord not using its working set loans on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO 494 Performance Manager Administrator Guide Performance Manager Data Cells PROC TYPEPRO L PROCTYPE This contains a hexadecimal representation of the type of process bit O interactive bit 1 batch bit 2 network bit 3 detached bit 4 subprocess for the current process subrecord on the local node for the current interval record Data Type NUMERIC Domains PRO PSWP WAITMET F SPMMMGWAIT Percentage of time that the CPU was idle and at least one disk device had either paging I O or swapping I O in progress for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU
237. SSES IN MWAITMET F MWAIT This contains a value representing the average number of processes in MUTEX or resource wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES IN PFWMET F PFW This contains a value representing the average number of processes in page fault wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 479 Performance Manager Data Cells PROCESSES IN SUSPMET F SUSP This contains a value representing the average number of processes in suspend wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES IN SUSPOMET F SUSPO This contains a value representing the average number of processes in outswapped suspend wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES_NEED_MORE_EXTENT Derived This contains a Boolean value representing the presence of one or more processes which need larger working set extents on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES NEED MORE WSMAX Derived This contains a Boolean value represent
238. STACKED of the items depicted up to 6 If the GREATER THAN keyword is specified without a value then 50 percent is assumed If the LESS THAN keyword is specified without a value then 90 percent is assumed 246 Performance Manager Administrator Guide GRAPH If the SELECT qualifier is present without a keyword then GREATER THAN 50 is assumed For example ADVISE PERFORMANCE GRAPH BEGINNING 10 ENDING 11 NODE YQUEM _ TYPE CPU U THRESHOLD 25 CPU M THRESHOLD 35 TOP CPU I THRESHOLD 45 _ SELECT GREATER OUTPUT REGIS PSPA I CREAGRAPHOUT PSPA Graph created file BADDOG CORREY WORK PSPA YQUEM CPU UTILIZATION REG 1 This command requests that three graphs be produced The CPU Utilization graph is produced if 50 percent or more of the data points exceed 25 percent CPU utilization The CPU MODES graph is produced if 50 percent or more of the data points exceed 35 percent CPU utilization The TOP CPU IMAGES graph is produced if 50 percent or more of the data points exceed 45 percent CPU utilization In this case only one graph is produced ADVISE PERFORMANCE GRAPH BEGINNING 10 ENDING 11 NODE YQUEM _ TYPE CPU U THRESHOLD 25 CPU M THRESHOLD 35 TOP CPU I THRESHOLD 15 _ SELECT GREATER OUTPUT REGIS 9PSPA I CREAGRAPHOUT PSPA Graph created file BADDOG CORREY WORK PSPA YQUEM CPU UTILIZATION REG 3 PSPA I CREAGRAPHOUT PSPA Graph created file BADDOG CORREY WORK PSPA YQUEM TOP CPU IMAGES REG 1
239. SWP CPU 162 180700 15491 65 5 56 6 58 7 MP Idle Synch Stack Kernel User Compat and I 0 Overlap Page Swap or Smp I 0 Multi CPU IO Wait Wait Wait Only I 0 Busy Paging Rates per second Page System Read Pages Write Free Modify Dzero Gvalid WritIn Hard Soft Faults Faults I Os Writen I 0s List List Faults Faults Prog Faults Faults I 0 Rates per second File I 0 Rates per second Direct Buffrd Lognam Mailbx Mailbx Window Window Split Erase File I 0s I 0s Trans Reads Writes Hits Turns I 0s I 0s Opens 74 Performance Manager Administrator Guide Tabular Report Sections Tabular Report YQUEM VAX 6000 440 PA Vx x Tuesday 26 JAN 2006 09 00 to 10 00 File Cache Attempt Rate per second Dir File File i i File File Data Quota Id Hdr Extent Quota Id Hdr Extent Directory Deadlock Blocking AST Functions Messages ENQ Dlock Dlock Total Total Wait NotQD Search Find Locks Resrcs System Communication Service Rates per second Data G Data G Msgs Msgs Snd Cr Reqst K Byte K Byte Buf Dsc Recvd Discd Sent Recvd Queued Queued User Command ADVISE PERF REPORT TAB SECT SUMM NODE YQUEM BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 10 00 00 00 0UT TAB SUM YQUEM Avg Process Memory Counts Average memory statistics provide page process and cache information These metrics are collected from OpenVMS performance statistics Proc Count The number of pro
240. Statistics 3 Enter the Help command or press PF2 to display the single node display keypad The following table shows the functions of the single node display keypad Invest KEYPAD DEFINITION igate Display for the Zoom SINGLE NODE DISPLAY in on Section Zoom Out to aZ Exit PSPA RT Multi Node Display Display CPU Utilization The CPU Utilization screen displays total utilization and the percentage of CPU utilization of each of the following processor modes m Interrupt stack m MP synchronization m Kernel mode m Executive mode m Supervisor mode m User mode Compatibility mode Chapter 11 Use the Character Cell Real time Display 379 Display Single Node Statistics Display Top Processes Statistics Process statistics are summarized and presented by the process key The process statistics are presented in descending order by the selected metric value To summarize the process statistics according to a different key m Press KP6 The available process keys are as follows TOP User metric TOP Image metric TOP Process metric TOP Account metric TOP PID metric To change the metric displayed m Press KP9 The available metrics are as follows CPU Disk IO Disk Thruput Soft Faults Hard Faults Buffered IO Direct IO WS pgs Private pages Global pages V A pages 380 Performance Manager Administrator Guide Display Process Information Display Top Device Statistics Device statistics are s
241. Step 1 Create a CSV file see page 559 Step 2 Create More CSV Files as Necessary see page 560 Step 3 Create a Single CSV File see page 560 Step 4 Send the CSV File to a Windows Machine see page 562 Step 5 Create the Graph in Excel see page 562 Step 1 Create a CSV file You need to create a CSV file that contains the custom graph data for first set of six CPUs The ADVISE PERFORM GRAPH command has a CUSTOM type option Use this command with the following options ADVISE PERFORM GRAPH NODE nodename OUT nodename cpus nnl nn2 CSV FORMAT CSV BEGIN dd mmm yyyy hh mm END dd mmm yyyy hh mm TYPE CUSTOM CPU METRICS P BUSY SELECTION cpuid cpuid cpuid cpuid cpuid cpuid Appendix F How You Graph Seven or More CPUs 559 Step 2 Create More CSV Files as Necessary For example the command for the first six CPUs of node RX8620 ADVISE PERFORM GRAPH NODE RX8620 OUT RX8620 CPUS 0 5 CSV FORMAT CSV BEGIN 15 JAN 2008 13 00 END 15 JAN 2008 13 10 TYPE CUSTOM CPU METRICS P BUSY SELECTION RX8620 0 RX8620 1 RX8620 2 RX8620 3 RX8620 4 RX8620 5 Step 2 Create More CSV Files as Necessary Repeat step 1 for each set of CPUs Ensure you add no more than six CPUs to each file Step 3 Create a Single CSV File You need to append the CPU data for the same collection interval to the existing record and keep the order of the CPUs in each record To create a single CSV file
242. TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK TOP OPERATION FILE X Derived This contains an index specifier identifying the hot file subrecord for the hottest file in terms of I O operations per second on the current disk on the local node for the current interval Data Type INDEX Domains DSK Target Domains FIL DISK TOP SPLIT IO FILE X Derived This contains an index specifier identifying the hot file subrecord for the hottest file in terms of split I O operations per second on the current disk on the local node for the current interval Data Type INDEX Domains DSK Target Domains FIL DYN EXPANSION COUNT Derived A count of the number of times nonpaged pool is increased for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY DYN MAXLEN Derived The maximum number of bytes in nonpaged pool for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY Appendix C Performance Manager Data Cells 445 Performance Manager Data Cells ENQUEUE LOCKS NOT QUEUED RATEMET F ENQNOTQD This contains the number of enqueue lock requests not queued per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP ENQUE_LOCKS_FORCED_TO_WAIT_RATEMET_F_ENQWAIT This contains the number of enqueue lock requests per second that had to enter the lock wait queue for the current interval for the
243. TD T95 RZ74 47 Same as previous TD T96 ESE52 1300 Same as previous TD T197 ESE56 1300 Same as previous TD T98 ESE58 1300 Same as previous TD T99 RZ27 51 Same as previous TD I1 RK06 530000 Number of bytes per second threshold for disk type Data cell DISK THRUPUT RATE THRESHOLD has this value if the current disk sub record is of this type TD I2 RK07 530000 Same as previous TD I3 RPO4 800000 Same as previous TD I4 RPO05 800000 Same as previous TD I5 RPO6 800000 Same as previous TD I6 RMO3 1200000 Same as previous TD I7 RPO0O7 1300000 Same as previous TD I8 RPO7HT 2200000 Same as previous TD 113 RM80 1000000 Same as previous TD 115 RMO05 1200000 Same as previous TD I20 RA80 800000 Same as previous TD 121 RA81 1500000 Same as previous TD I22 RA60 1300000 Same as previous TD I25 RD51 300000 Same as previous 144 Performance Manager Administrator Guide Components of Rules Name Default Description Value TD 127 RD52 530000 Same as previous TD 128 RD53 530000 Same as previous TD I30 RA82 1600000 Same as previous TD I31 RD31 500000 Same as previous TD I32 RD54 530000 Same as previous TD I34 RRD50 120000 Same as previous TD I36 RX33 400000 Same as previous TD I37 RX18 400000 Same as previous TD I38 RA70 1000000 Same as previous TD I39 RA90 1900000 Same as previous TD 140 RD32 500000 Same as previous TD I42 RX35 400000 Same as previous TD I43 RF30 900000 Same as previous TD 144 R
244. Target Domains CFG TOP BUFIO PROCESS X Derived This contains an index pointing to the process subrecord that has the highest buffered I O operations rate on the local node for the current interval Data Type INDEX Appendix C Performance Manager Data Cells 521 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP COM PROC BPRI Derived This contains a value representing the base priority of the process subrecord with the most time in the COM state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TOP COM PROC BPRI A Derived This contains a value representing the base priority of the process subrecord with the most time in the COM state If PRIORITY LOCKOUT is true 1 0 this cell will contain the base priority of the process which is probably unable to utilize the CPU because of a priority lockout and pointed to by the cell TOP COM PROC X A Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TOP COM PROC X Derived This contains an index pointing to the process subrecord with the most time spent in the scheduler computable state on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO TOP COM PROC X A Derived This contains an index pointing to the process subrecord with the highest percentage of time being computable If PRIORITY
245. This command produced two of three graphs because threshold quantity for the last graph was lowered ADVISE PERFORMANCE GRAPH BEGINNING 10 ENDING 11 NODE YQUEM _ TYPE CPU U THRESHOLD 25 CPU M THRESHOLD 35 TOP CPU I THRESHOLD 15 SELECT GREATER 90 0UTPUT REGIS The previous command generated none of the graphs because in all cases 90 percent of the graph points did not exceed the specified thresholds STACK NOSTACK Stacks the values for each category on the graph Use NOSTACK to overlay the values on the graph ReGIS graphs using NOSTACK may cause some occlusion if you do not specify FORMAT ReGIS CHARACTERISTICS LINE also If you are requesting a series of graphs in one command you can override the NO STACK qualifier by specifying the NO STACK keyword following each graph type See Chapter 4 for an illustration of the use of the NOSTACK qualifier and for additional information about default behavior Chapter 7 Use Command Mode Commands 247 PIE CHART TYPE graph_type NO STACK Y_AXIS_MAXIMUME n THRESHOLD m TITLE string ALL GRAPHS NO STACK THRESHOLD m Y AXIS MAXIMUMz n CUSTOM see TYPE CUSTOM in Chapter 6 PROMPT Specifies which of the graphs you want generated The PROMPT keyword specifies that the Performance Manager prompt you for the graph types and custom metrics Using PROMPT has the advantage of allowing an interactive user the ability to preview any predefined or
246. UNTVPAR F SRPCOUNTV The value of the SYSGEN parameter SRPCOUNTV which establishes the upper limit to which SRPCOUNT can be automatically increased by the system for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 516 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN SRPSIZEPAR F SRPSIZE The value of the SYSGEN parameter SRPSIZE which indicates the size in bytes of the small request packets for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN SWPALLOCINCPAR F SWPALLOCINC The value of the SYSGEN parameter SWPALLOCINC which sets the swap file allocation increment value in blocks used to backup swap file allocation space in the swap or page file for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN SWPOUTPGCNTPAR F SWPOUTPGCNT The value of the SYSGEN parameter SWPOUTPGCNT which defines the minimum number of pages pagelets to which the swapper should attempt to reduce a process before swapping it out for the local node for
247. Us for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP ANYIO BUSYMET F SPMIOBUSY Percentage of time that at least one disk device was busy for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP ANY DISK FULL Derived This contains a Boolean value zero or one where one truth represents the fact that the percentage of free space on any disk is less than or equal to the minimum disk free space percentage threshold TD MIN DSKSPC PCT The data cell refers to the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP ANY DISK OVER QL THRESHOLD Derived This contains a Boolean value zero or one where one truth represents the fact that the queue length on any disk is greater than or equal to the maximum disk queue length threshold TD DISK QL MAX The data cell refers to the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 413 Performance Manager Data Cells ANY DISK OVER THRESHOLD Derived This contains a Boolean value zero or one where one truth represents the fact that the operations rate on any disk is greater than or equal to the threshold for that disk ty
248. VESTIGATE commands to control the display and its characteristics INVESTIGATE Command Options The INVESTIGATE commands are shown in the following table Command INVESTIGATE CPU DISPLAY INVESTIGATE IO DISPLAY INVESTIGATE LOAD BALANCE DISPLAY INVESTIGATE MEMORY DISPLAY INVESTIGATE SYSTEM OVERVIEW DISPLAY Function Displays CPU statistics Displays I O statistics Displays load balance statistics Displays memory statistics Displays system overview statistics Chapter 11 Use the Character Cell Real time Display 391 Evaluate Performance Using the Investigate Displays The Load Balance display is the only Investigate display available for both ReGIS and DEC CRT terminals It contains the same information in both cases On ReGIS terminals it is a Kiviat graph and for DEC CRT terminals it is a bar graph The System Overview is the default Investigate display for ReGIS compatible terminals and for non ReGIS or DEC CRT terminals the default is the Load Balance display The following illustrations show an example of the ReGIS version of the system Overview display an example of the ReGIS Load Balance display Kiviat and an example of the DEC CRT ANSI version of the Load Balance display Additional displays show memory I O and CPU statistics and require a ReGIS compatible terminal There is also an example of the Memory display an example of the I O display and an example of the CPU display
249. WPRO V SSS PFW This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the page fault wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWASTPRO V RSN ASTWAIT This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWAST mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 490 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS WAS IN RWBRKPRO V RSN BRKTHRU This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWBRK mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWCLUPRO V RSN CLU This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWCLU mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWIMGPRO V RSN IACLOCK This contains a Boolean value where 1 represents the fact t
250. YPEMAG_A_HWTYPE This contains hardware type string of the node which serves the tape drive to the cluster for the current tape record in TAPE domain Data Type STRING Domains TAP 520 Performance Manager Administrator Guide Performance Manager Data Cells TAPE SERVER NODENAMEMAG A NODENAME This contains nodename string of the node which serves the tape drive to the cluster for the current tape record in TAPE domain Data Type STRING Domains TAP TERMINAL IO Derived This contains a value representing the sum of terminal operations rate to all communications terminals on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TICKER IS ON Derived The value of the SYSGEN parameter MMG CTLFLAGS has bit zero set for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP TIME Derived This contains the date and time associated with the current interval for either the local node or the cluster wide I O data cells Data Type TIME Domains COM CFG CLU CPU DSK FIL LOC PRO TAP TOP BDTW SCS NODE X Derived This contains an index pointing to the configuration subrecord for the SCS node with the most BDT waits on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP
251. _F_NP_FREE This contains the number of free bytes in non paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BYTES_FREE_IN_PAGED_POOLMET_F_PG_FREE This contains the number of free bytes in paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BYTES_IN_NONPAGED_POOLMET_F_NP_POOL_MAX This contains the number of bytes in non paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BYTES_IN_PAGED_POOLMET_F_PG_POOL_MAX This contains the number of bytes in paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 418 Performance Manager Administrator Guide Performance Manager Data Cells CACHE FREEMET F CACHE FREE This contains the number of free pages in the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE MAXIMUMMET F CACHE MAXI This contains the maximum number of pages SPTEs in the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE MISSES LT33MET F CACHE MISS LT33 The number of read operations with a block size less than 33 that bypassed the XFC for the current interval for the local node Data Type NUMERIC Domain
252. a files are created by the Performance Manager CPD data collector one for each node in the cluster each day The filename has the following format PSDCS DATABASE PSDC nodename yyyymmmdd CPD Data cell A data cell is the basic unit of data used to create analysis reports This data is either retrieved directly from a field in a subrecord of a daily data record or derived from it The data cell is typically used as a variable in a rule expression Data collection error log The data collection error log is an ASCII file common to the entire cluster system called PSDC DATABASE PSDC DC LOG Errors that occur during data collection are recorded in this file 566 Performance Manager Administrator Guide Data collection schedule The data collection schedule is a user defined schedule by which the Performance Agent determines when to record data and what data to record for each node in the cluster Data collection synchronization Data files Data record Database directory Dates file Device Direct I O Data collection synchronization is a method used to correlate intervals measured on different nodes within a cluster system those intervals must represent the same real time to make analysis effective Data files are the files containing performance data from which reports and graphs are constructed The data record contains performance data written by the data collector The CPD data collector writes one data record for ea
253. ac This column provides the maximum number of virtual address space pages for any one process for any one recording example m Max WSsize This column provides the maximum working set size private pages global pages of any one process for any one recording example Chapter 3 Evaluate Performance in Detail 49 Performance Evaluation Report m Ave WSsize This column provides the average working set size private pages global pages for the process For more than one process the value is an average weighted by the processes uptime m Ave WSlist This column provides the average working set list size for the process For more than one process the value is an average weighted by the processes uptime m WS Default This column provides the WSDEFAULT value for the process For more than one process the value is an average weighted by the processes uptime WS Quota This column provides the WSQUOTA value for the process For more than one process the value is an average weighted by the processes uptime m WS Extent This column provides the WSEXTENT value for the process For more than one process the value is an average weighted by the processes uptime m Uptime min This column provides the total amount of wall clock time during which the process es were resident on the system in minutes Interpret Pool Statistics The pool statistics follow the process statistics for each node in a cluster system To display only the pool s
254. ac WSsize WSsize WSlist Default Quota Extent SENDLOSKY COPY ARROYO BHAT FORD QUANG VERRIER User Command ADVISE PERF REPORT PERF NODE YQUEM BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 10 00 00 00 INCLUDE PROC PROC FOCU MEM OUT EXAMPLE 3 MEM INFO The following statements describe the columns in the previous example m Primary Secondary Key This column lists the primary and secondary key identifiers for the process detail lines The keys may be Mode Username Imagename Processname Account UIC group Workloadname or PID m SftFlt Sec This column provides the average soft pagefault rate per second for the process or processes m HrdFit Sec This column provides the average hard pagefault rate per second for the process or processes m Ave Private This column provides the average number of private workingset pages for the process For more than one process the value is an average weighted by the processes uptime m Ave Global This column provides the average number of global workingset pages for the process For more than one process the value is an average weighted by the processes uptime m Ave VA Spac This column provides the average number of process virtual pages for the process For more than one process the value is an average weighted by the processes uptime m Obs PFW This column is set to a Y if the process was observed by the data collector in the pagefault wait scheduler state PFW m Max VA Sp
255. age activations as a count for the process or processes Cpu time Min This column provides the amount of CPU time consumed by the process or processes in minutes Chapter 3 Evaluate Performance in Detail 43 Performance Evaluation Report m DirI O Sec This column provides the average Direct I O rate per second for the process or processes m BufI O Sec This column provides the average Buffered I O rate per second for the process or processes m SftFlt Sec This column provides the average soft pagefault rate per second for the process or processes m HrdFit Sec This column provides the average hard pagefault rate per second for the process or processes m Ave WSsize This column provides the average working set size private pages global pages for the process For more than one process the value is an average weighted by the processes uptime m Max WSsize This column provides the maximum working set size private pages global pages of any one process for any one recording example m Ave Global This column provides the average number of global workingset pages for the process For more than one process the value is an average weighted by the processes uptime m WS Default This column provides the WSDEFAULT value for the process For more than one process the value is an average weighted by the processes uptime m WS Quota This column provides the WSQUOTA value for the process For more than one process the
256. aged by other nodes 5 Average number of ENQ lock requests per second that had to wait in the wait queue 6 Value of the SYSGEN parameter LOCKDIRWT on each respective node 7 Deadlock detections per second during the reporting period 8 Deadlock searches per second during the reporting period Interpret cluster wide CI NI and Adapter Statistics The CI NI and adapter statistics follow the cluster wide Lock statistics To display only the CI NI and adapter statistics section of the Performance Evaluation Report use the following qualifier INCLUDE CI NI AND ADAPTER STATISTICS To disable the CI NI and adapter statistics display from the report use the following qualifier INCLUDE NOCI NI AND ADAPTER STATISTICS Note The phrase CI NI and adapter statistics is seen in the Performance Evaluation Report to describe cluster interconnect statistics in general CI hardware is not supported on HP Integrity Servers On a cluster of only Integrity Servers you will see only NI adapters Chapter 3 Evaluate Performance in Detail 55 Performance Evaluation Report Performance CLUSTER Page 7 Evaluation PA Vx x Wednesday 14 JAN 2006 00 00 to 12 16 Data Mess Disk Disk Transfers Component grams Operations KB Thruput KB Thruput total total DEFEND CIBCA A SUPPLY C1780 AGRIC CIBCA A SUPPLY BCI750 TAYLOR CIBCA A NODE2 C1780 NODE2 C1780 Un 8 6 1 0 0 3 8 6 6 2 2 8 9 0 0 8 2 0 0 cOoOcococococooco
257. ager Administrator Guide Qualifiers Examples ADVISE PERFORMANCE COMPILE During the archiving process the auxiliary rules can be applied to the data being processed resulting in rules records being added to the archive files Later processing of the archived data allows the rules records to be graphed reported on or dumped Optionally set AUTO AUGMENT in the parameters file and specify the file specification for the auxiliary knowledge base The Performance Manager software uses your rules to augment the Performance Manager factory rules For information and additional examples on how to create an auxiliary rules file see the chapter Customize the Knowledge Base see page 135 RULES file spec Specifies the name of the auxiliary knowledge base output file By default the Performance Manager rules compiler creates an output file in the default directory with the file name of the specified input file and file type of KB ADVISE PERFORMANCE COMPILE MY SITE The Performance Manager reads the MY SITE VPR file and compiles the rules for subsequent use The Performance Manager creates a file named MY SITE KB in the default directory To use the compiled rules in conjunction with Performance Manager factory rules type the following command ADVISE PERFORMANCE REPORT ANALYSIS RULES MY SITE Or alternately use the Parameter Edit Utility to automatically use your compiled rules ADVISE EDIT PSDC EDIT SET AUTO
258. ains the number of bytes the smallest block in paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SMALL BLKS FREE NONPAGED POOLMET F NP FREE LEQU 32 This contains the number of free blocks less than or equal to 32 bytes in non paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 499 Performance Manager Data Cells SMALL BLKS FREE PAGED POOLMET F PG FREE LEQU 32S This contains the number of free blocks less than or equal to 32 in paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SOFT_FAULT_RATE Derived Average number of soft pagefaults per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SOFT_FAULT_SCALING Derived This contains a value representing scaling factor for the soft page fault rate for the local node The value is obtained from the threshold TD_SOFT_FAULT_SCALING_n where n is the hardware model number of the local node By default if the local node is a VAX 11 780 the value would be 1 0 The value of this data cell can be modified using a threshold construct Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SPLIT_IO_RATEMET_F_SPLIT Average split I O rate per second for the local nod
259. al and outgoing enqueue and local and outgoing converted enqueue operations Dir Op Incomg sec This metric is the Directory Operations Incoming per second This is the number of lock directory operations per second being requested of the host node R Orig Lck Act sec This is the number of locking operations performed on the host node on behalf of other nodes in the cluster It is equal to the sum of incoming enqueue incoming converted enqueue and incoming dequeue operations 98 Performance Manager Administrator Guide Tabular Report Sections Cluster Disk and Server Statistics with By Node Breakout The Performance software reports disk statistics By Cluster and By Node Cluster statistics represent the total of all disk I O for a cluster By Node statistics represent each node s contribution to the cluster I O load The following example shows disk statistics by cluster Tabular Report CLUSTER PA Vx x Tuesday 26 JAN 2006 09 00 to 09 20 SEAS kk Final Statistics SASSO I k kkk aaatadabaabaabaeaabarek 33 Data Analyzed from 26 JAN 2006 09 00 00 00 to 26 JAN 2006 09 20 00 00 FAAARA Paging Swping Rate Time Queue Space ms Length Used DSA1 DSA10 DSA111 DSA12 DSA13 DSA14 YQUEM DFSC7104 YQUEM DFSC7105 User Command ADVISE PERF REPORT TABULAR BYCLUSTER BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 09 20 00 00 0UT BYCLU TAB SECTION DISK The way the Performance software counts a node s contribut
260. al node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP OUTGOING_DEADLOCK_MESSAGE_RATEMET_F_DLCKMSGS_OUT This contains the number of outgoing deadlock detection messages per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP OUTGOING_DIRECTORY_FUNCT_RATEMET_F_DIR_OUT This contains the number of outgoing directory operations per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP OUTGOING_LOCK_CONVERSION_RATEMET_F_ENQCVT_OUT This contains the number of outgoing enqueue lock conversion requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP OUTGOING_LOCK_DEQUEUE_RATEMET_F_DEQ_OUT This contains the number of outgoing dequeue lock requests per second for the current interval for the local node Data Type NUMERIC Appendix C Performance Manager Data Cells 473 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP OUTGOING LOCK ENQUEUE RATEMET F ENQNEW OUT This contains the number of new outgoing enqueue lock requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP OUT SWAP RATEMET F OSWPCNT Average outswap rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK F
261. al time Display 347 How You Edit the Panel Instruments You can enter any of the following values Maximum Data Value Specifies the maximum value of the graph s scale The default value is 100 Minimum Data Value Specifies the minimum value of the graph s scale The default value is O Low Threshold Enables and specifies a value line to be displayed on the graph Data below this line is displayed in the patterns and colors you set for low threshold High Threshold Enables and specifies a value line to be displayed on the graph Data above this line is displayed in the patterns and colors you set for high threshold Peak Hold Units Enables peak hold and specifies the number of units of time the peak maximum value attained by the metric is held in the display This peak value indicator is displayed in the patterns and colors you set for the peak hold Strip Chart Automatic Scaling Maximum Data Value 100 00 Minimum Data Value 0 Number of Time Units L High Threshold 30 00 348 Performance Manager Administrator Guide How You Edit the Panel Instruments The Strip Chart Range and Thresholds dialog lets you do the following actions m Set automatic scaling m Set maximum and minimum data values m Set number of time units along the x axis m Set low and high thresholds You can enter any of the following values Automatic Scaling Specifies that the height of the chart is to be dynamically adjust
262. anager Features You can also request a Performance Evaluation Report or Tabular Report to help you determine improved or degraded system performance You can also request histograms that consist of chronological charts that show peak resource use You can select a specific set of items disks or processes for reporting The ADVISE PERFORMANCE REPORT command invokes analysis of performance data to generate Performance Manager reports You can specify five types of reports m Analysis Report Consists of conclusions for each node and includes cluster wide conclusions for a clustered system You can request the conditions that caused a rule to fire and the supporting evidence m Brief Analysis Report Is a brief version of the analysis report It includes a one line synopsis of each rule firing Histograms Consists of chronological charts that show peak resource use m Performance Evaluation Report Consists of metrics for measuring performance improvement or degradation Use these metrics to evaluate the recommendations made by Performance Manager and to measure results against baseline models Tabular Report Provides a consolidated summary of some of the performance metrics related to system wide activity process activity and disk activity Generating Daily Reports Automatically You can generate daily reports automatically by submitting a Performance Manager batch command procedure at night The Performance Manager software kit con
263. anager Real time Process Display Multiple Processes You can scroll through the process instances as appear in the previous illustration by using NextScreen and PrevScreen keys If there is a particular PID user name or image name that you want to lock the display on enter the USERNAME IMAGENAME or PID command to do so For example PSRT5 USER HOFFMAN The last section of the display is the volumes section also appearing in the earlier illustration showing Performance Manager Real time Process Display Multiple Processes where disk volumes and I O rates appear These represent the top disks that the processes use and are list up to the top five disk volumes recorded by the main collector By pressing KP3 you can view the top files being used by the processes The top volumes and files data is available only when the MODE DISKFILE command and the default collection definition is used COLLECTION CPD Press PF2 to view the keypad for Process displays The following illustration shows the functions of the Process display keypad Resource invest Display igate KEYPAD DEFINITION Display for the PROCESS DISPLAY Process Next and Prev Screen Next Next Togg Top Scroll the images Top Node Volume if listed in the Process or File display mum Out to un aZ Exit PSPA RT the mum Node Display Chapter 11 Use the Character Cell Real time Display 383 Display Disk Information Display Disk Information Performance
264. are calculates By Cluster disk statistics Disk Node Relationship Calculation of Total Cluster Disk Statistics One or more nodes with direct Add I O data from node s with direct access and any number of access to disk nodes with remote access No nodes with direct access in Add I O data from all nodes with remote the collection specification and access only one or more nodes with remote access Depending upon the number of nodes with direct and remote access to the disk computing By Node I O disk statistics is complex and sometimes not possible to calculate for all nodes This is because it is not always possible to distinguish each node s contribution to the total Cluster I O rate When there is only one node with direct access to a disk and any number of nodes with remote access the By Node contribution of the node with direct access is calculated as follows Subtract the I O statistics of any nodes with remote access from the I O statistics of the node with direct access The By Node contribution of nodes with remote access to the disk is the I O statistic for each remote node In the case of more than one node with direct access to a disk and more than one node with remote access calculation of each node s contribution is not possible because there is no way to distinguish which node with direct access performed the I O operations for which nodes with remote access The following table shows a summary of how the software calcu
265. as in the computable state most during the interval 7 Base priority of the process that was in the computable state most during the interval 8 Beginning time of the interval in which the condition occurred Chapter 2 Analyze Performance 25 Analysis Reports In the previous CPU Rule analysis report example the average number of processes in either COM or COMO state is five or greater on four occasions with the actual number of COM COMO processes ranging from 8 to 20 These blocked computable processes for users Jones Doe Mack and Hall each have a base priority of 4 Other processes with a base priority of 8 for users Smith John Tom and Jerry prevent the other computable processes from executing because of their elevated base priority thereby creating the LOCKOUT condition O Rule Analysis Report The following example shows an I O Rule analysis report Full Analysis SUPPLY VAX 11 780 PA Vx x Saturday 01 0CT 2006 00 00 to 23 59 CONCLUSION 3 I0060 Swapping or modified page writing is creating an excessive load on the noted disk This may be a memory related problem however if the swapping file is on a shared system disk the situation can be improved by moving it to a less utilized disk Total number of samples supporting this conclusion 4 CONDITIONS 1 ANY DISK OVER QL THRESHOLD EQ 1 00 2 PERCENT CPU TIME IN FILE SYSTEM LT 30 00 3 EXEC LT 35 00 4 DISK SWAPPING IO RATE DISK OVER QL THRESHOL
266. ast interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PIXSCANPAR F PIXSCAN The value of the SYSGEN parameter PIXSCAN which determines the maximum number of processes to scan for priority boosting for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN POOLCHECKPAR F POOLCHECK The value of the SYSGEN parameter POOLCHECK which enables a reserved debugging aid in locating problems within OpenVMS data structures by verifying memory allocations and deallocations for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Appendix C Performance Manager Data Cells 513 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PQL DWSDEFAULTPAR F PQL DWSDEFAULT The value of the SYSGEN parameter PQL DWSDEFAULT which sets the default working set size for a process created by the Create Process CREPRC system service or the DCL command RUN Process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PQL DWSEXTENTPAR F PQL DWSEXTENT T
267. at incurred the highest read I O rates The metric graphed is FILE READIO TYPE TOP READ VOLUMES Top Read Operations Disk Volume Plots the five plus Other Volumes disk volumes that incurred the highest read I O rates The metric graphed is READIO TYPE TOP RESIDENT IMAGES Most Resident Images Plots the top five plus Other Images images most resident on the system by image name The metric graphed is RESIDENCE TYPE TOP RESIDENT USERS Most Resident Users Plots the top five plus Other Users users most resident on the system by user name Note that each subprocess adds to the residence for the parent process s user name The metric graphed is RESIDENCE TYPE TOP RESIDENT WORKLOADS Most Resident Workloads Plots the top five plus Other Workloads workloads most resident on the system by workload name The metric graphed is RESIDENCE TYPE TOP RESOURCE RULE OCC Top Resource Rule Occurrences Plots the five plus Other resource rule identifiers that fired as a rate per hour The metric graphed is RESOURCE OCCURRENCES and is available only from history data 226 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE TOP RESPONSE TIME DISKS Top Response Time Disk Device Plots the five plus Other Disks disk devices that incurred the highest response times The metric graphed is D RESPONSETIME TYPE TOP RESPONSE TIME FILES Top Response Time Files Plots the five plus
268. ault panels distributed with the Performance Manager gt lt neal time v3 0 0610 OLX File Panel Help instrument Panel Directory Buffered 1 0 mage CPU mage CPU MODE DISK FILE Memory Allocation Navigate Within the Default Panels The Real time Display default instrument panels facilitate a progressive disclosure style of investigation in which increasingly detailed data is presented to you You inquire about a particular resource by double clicking on the resource name displayed This process is known as aunching You are then presented with panels of information about the use of the resource the top users or processes requiring the resource for example If you need more detail you can launch additional panels by double clicking on a field in the instrument You can determine if the panel launch capability of the Real time Display is enabled for an instrument by moving the pointer over the graph or label displayed within the instrument The pointer changes to a plus sign if more information is available for this performance metric For example the four graphs displayed in the System Overview panel are entry points for disclosing additional information on CPU utilization page faulting and disk I O activity Moving the pointer into the graphs causes the pointer to change shape informing you that a double click operation here causes a new panel to be displayed To close a panel m Choose the Close menu item from the F
269. ay 26 JAN 2006 09 00 to 10 00 The following table summarizes the workload characteristics on a per image activation basis Note that values would be zeros if total number of image activations is zero PrimaryKey Avg Avg Avg Avg Workld of Soft Hard Direct Buff d Avg Images Secondary Key processes WSiz flts flts I0 I0 Cputim activ inact image image image image image EDITORS 4755 2087 2 17 5 317 7 947 0 7 59 NETWORK 1394 47 3 2 2 24 5 189 7 1 55 OTHER 1741 1879 0 27 7 1101 8 3228 2 42 17 SYSMAN 865 1967 6 22 1 241 0 661 3 31 10 UTILITIES 1952 290 4 6 9 77 6 316 0 1 40 The following statement is keyed to the column in the previous example The Primary Key Workld indicates the workload names associated with the specified workload family View Process Statistics with a Generalized Set of Metrics The following report example illustrates the process statistics section with a focus on CPU memory and IO data primarily presented as rates Some UAF parameters are also provided This format of the report is obtained by using the qualifier PROCESS STATISTICS FOCUS GENERAL 42 Performance Manager Administrator Guide Performance Evaluation Report To view the report with a different orientation provide your choice of sort keys For example the following syntax presents the statistics by workload family with a breakdown by image PROCESS STATISTICS FOCUS GENERAL PRIMARY KEY WORKLOAD SECONDARY KEY IMAGENAME
270. be used to wildcard either the group or user field of the specified UICs Specify FILTER PIDS to export all process records with the PID matching any of the specified PIDs Specify FILTER VOLUMENAMES to export all disk records with the volumename matching any of the specified strings Do not specify any trailing colon Chapter 6 Performance Manager Commands 185 ADVISE PERFORMANCE EXPORT Keyword Description DEVICENAMES string Specify FILTER DEVICENAMES to export all disk records with the devicename matching any of the specified strings Do not specify any trailing colon HISTORY history descriptor name Allows you to select history data from the Performance Manager database By default daily CPD files are processed However by specifying the name of a history file descriptor you can select historical data instead You must have previously defined the descriptor name in the parameters file and have used the archiving facility to create the history files Use the DCL command ADVISE EDIT to start the parameters editor From the utility you can ADD DELETE MODIFY and SHOW history file descriptors If you want to use history data instead of daily data use the HISTORY qualifier instead of the COLLECTION_DEFINITION qualifier These two qualifiers are mutually exclusive INTERVAL seconds Specifies the elapsed time to be summarized in an output record Its minimum value is that of the performance data file
271. ber of intvls w expansns 0 Allocation XRPCOUNT 962048 Virtual Alloc xRPCOUNTV 4810240 Virtual I 0 Cache Average Total Size Average Read I O Average Free Average Read Hit Average in Use Average Write I 0 Maximum Size SPTEs I 0 Bypassing the Cache Average Files Retained Cache Effectiveness User Command ADVISE PERF REP PERF INCLU POOL BE 19 MAR 2006 23 00 00 00 NOD MUMMS OUT A A Interpret CPU Mode Statistics The CPU mode statistics follow the pool statistics for each node in a cluster system To display only the CPU mode statistics section of the Performance Evaluation Report use the qualifier INCLUDE MODE STATISTICS To disable the CPU mode statistics display from the report use the qualifier INCLUDE NOMODE STATISTICS Chapter 3 Evaluate Performance in Detail 51 Performance Evaluation Report The following example shows the CPU mode statistics section of the Performance Evaluation report Performance SUPPLY VAX 11 78 Page 4 Evaluation PA Vx x Wednesday 14 JAN 2006 00 00 to 12 16 The following table gives the average percent of time in each of the various CPU modes for each active processor in the local node Samples is the record count contributing to the summary line The following statements are keyed to the previous example Physical processor identification Percentage of time in kernel mode for this physical processor Percentage of time in executive mode for this physical processor
272. ble menu item in the submenu Modify the Instruments To modify the instruments 1 Pull down the Edit menu choose the Build Mode menu item and click the Enable submenu item 2 Select an instrument 3 Pull down the Edit menu choose the Modify menu item and select a menu item in the submenu 346 Performance Manager Administrator Guide How You Edit the Panel Instruments This Modify submenu lets you do the following actions m Set ranges and thresholds m Set patterns and colors m Modify parts The contents of the Modify submenu varies depending on the type of instrument selected These submenu items are also available in a pop up menu when you press MB3 in the window work area Set Ranges and Thresholds Clicking on the Ranges and Thresholds menu item Performance Manager displays the Ranges and Thresholds dialog for the selected instrument Depending on your selected instrument Performance Manager displays one of the following dialoges m Bar Graph Range and Thresholds dialog m Strip Chart Range and Thresholds dialog The following example shows the Bar Graph Range and Thresholds screen Maximum Data Value Minimum Data Value CJ Low Threshold CJ High Threshold CJ Peak Hold Units The Bar Graph Range and Thresholds dialog lets you do the following actions m Set maximum and minimum data values m Set low and high thresholds m Set peak hold units Chapter 10 Customize the DECwindows Motif Re
273. bstraction of a system focusing on high level performance characteristics MODEL TRANSACTIONS Model transactions are a default workload family defined in the parameters file It can be used to characterize workloads in the transaction class for modeling MODEL USERGROUPS Modeling MSCP OpenVMS Cluster OTHER Model usergroups is a default workload family defined in the parameters file It can be used to characterize workloads in the user group class for modeling Modeling is the process of gathering organizing and evaluating principal components of a system and the ways in which they interact for the purposes of understanding and predicting system behavior MSCP Mass Storage Control Protocol A software protocol used to communicate between a VAX or Alpha processor and a disk controller such as an HSC An OpenVMS Cluster is a highly integrated organization of AlphaServer and HP Integrity server system or VAX and HP AlphaServer system applications operating systems and storage devices All workload families have the catch all workload OTHER to absorb process data that does not match the selection criteria of any defined workload 570 Performance Manager Administrator Guide Parameters file Performance Agent PSDC PARAMS DAT is the parameters file resides in the PSDC DATABASE area and serves as a repository for workload characterizations history file descriptors and other Performance Manager parameters The Performa
274. ccess to the data established for the target node using proxies or network objects Refer to the Performance Agent Administrator Guide for more information To use the Real time feature with DECnet Phase V when Phase V Node Synonyms are not defined you will need to create a node name translation file and use the DNS NAME qualifier For more information on this logical name PSPASDNS NAMES and the translation file refer to the appendix Performance Manager Logical Names see page 405 This section contains the following topics Start the Real time Display see page 321 Control the Real time Display see page 322 Navigate Within the Default Panels see page 323 Use the Panel Commands Menu see page 324 Default Panel Descriptions see page 324 Review Data in Playback Mode see page 337 Set the Thresholds and Ranges see page 338 Change the Colors and Patterns see page 339 Start the Real time Display To start the Real time Display m Enter the following command ADVISE PERFORMANCE DISPLAY WINDOW MODE NETWORK DISKFILE NODE nodename For more information about command syntax see the chapter Performance Manager Commands see page 175 Chapter 9 Use the DECwindows Motif Real time Display 321 Control the Real time Display Note The output for the Real time Display must be directed to your display using the DCL SET DISPLAY command For more information on starting Perfor
275. cesses in the system including SWAPPER Balset Count The number of processes resident in the balance set Chapter 3 Evaluate Performance in Detail 75 Tabular Report Sections Memory Utilization Free Pages The free page list size that is based upon the average number of pages in the free list for the reporting interval Modify Pages The modified page list size that represents the average number of modified pages for the reporting interval This section reports memory utilization for the reporting interval Total MEMutl The percentage of total available memory that is being utilized computed as total system memory free pages Total Memory Paged MEMutl The percentage of pageable memory utilized in the interval computed as paged memory free pages Paged Memory User MEMutl The percentage of user memory being utilized in the interval computed as user memory free pages User Memory Modify MEMutl The percentage of modified memory being utilized computed as modify pages user memory Avg Mem CPU Queues This section reports the number of times processes were waiting for memory or CPU Mem The average number of processes waiting for available memory Equivalent to the count of processes in the computable outswapped queue COMO CPU The average number of processes waiting for the CPU There is a queue if this number is greater than one 1 the process that would have run i
276. ch two minute interval The database directory is a directory located on a permanently mounted disk accessible to every node in the cluster The data collection process writes the daily data files to this area Performance Manager software references this area via the system wide logical name PSDC DATABASE The dates file contains a list of dates used to select data A device is a piece of hardware in the computer system It performs measurable units of work The number of direct I O operations performed per second This illustration is tallied at the QIO application interface layer Disk branch by source Disk I O Dump report In a modeling context disk branch by source is the probable distribution of load across disks by originating CPU for a workload or a transaction class Probabilities must sum to one This data is included in the model input file Disk I O is the number of I O operations per second for the device This illustration is tallied at the physical device driver layer A dump report contains formatted output of data fields for each record of a Performance Manager daily data file or history file Glossary 567 Evidence Factory rules Family name Family type File type Granularity Hard page fault Evidence supports lines of performance data displayed in an analysis report Factory rules are the performance rules supplied with the Performance Manager in the PSPASEXAMPLES area in the file named PSP
277. chooses a best fit value for x points so that the time period represented by each point is even As the value of X POINTS increases spikes and valleys become more defined and the graph has a higher resolution A low number of X POINTS produces a smoother graph because the graphing facility averages any additional data points within the time frame requested Consider the time frame of a particular graph request when you determine the value of X POINTS For example over a 12 hour span the Performance Manager records statistics 360 times every 2 minutes If the value of X POINTS is 24 the graphing facility averages every 15 data records or 30 minutes and produces a graph with smooth flow If the value of X POINTS is 72 the graphing facility averages every 5 data records or 10 minutes and produces a graph with valleys and spikes Use the WIDTH keyword to specify the column width of the ANSI graph output Valid widths range from 40 to 132 columns If you do not specify the WIDTH qualifier the Performance Manager uses the terminal width setting When you specify the OUTPUT qualifier or generate the graph under batch the width of the graph is 132 columns Use the HEIGHT keyword to specify the graph height of the ANSI graph output Valid heights are from 20 to 60 lines If you do not specify HEIGHT the Performance Manager uses the terminal page length setting When you use the OUTPUT qualifier or generate the graph under batch the heig
278. ck statistics section of the Performance Evaluation Report use this qualifier INCLUDE LOCK_STATISTICS To disable the lock statistics display from the report use this qualifier INCLUDE NOLOCK_STATISTICS The following example shows the lock statistics section of the Performance Evaluation Report Performance CLUSTER Page 14 Evaluation PA Vx x Wednesday 14 JAN 2006 00 00 to 12 16 The following table gives a summary of the average amount of lock traffic per second in the cluster Local Incoming Outgoing Waiting LOCK Deadlk Deadlk Node ENQ CVT DEQ ENQ CVT DEQ ENQ CVT DEQ locks DIRWT find search NODEOl 24 23 24 13 3 13 Total 24 23 24 13 3 13 0 1 54 Performance Manager Administrator Guide Performance Evaluation Report The following statements are keyed to the columns in previous example 1 Each node in the cluster 2 Average enqueue ENQ conversion CVT and dequeue DEQ lock requests per second for locks that are managed by the node requesting the lock An enqueue lock request queues a new lock resource A conversion lock request occurs when a lock of one mode has already been granted and a lock request to change the lock mode is to be granted A dequeue lock request releases the granted lock 3 Average lock requests per second for locks that are managed by the local node but originate on other nodes 4 Average lock requests per second for locks that originate on a local node but are man
279. coco eoooooocoeo0oo DD HOOD DDTDDTOOTWAOA WOO 2 oococn uocosc cQ eG D ora dad amp DDO GD DQ QD Do DOT dada User Command ADVISE PERF REPORT PERF NODE YQUEM BEG 26 JAN 1997 09 00 00 00 END 26 JAN 1997 10 00 00 00 INCLUDE PROC PROC FOCU IO OUT EXAMPLE 3 IO INFO The following statements describe the columns in the previous example m Primary Secondary Key This column lists the primary and secondary key identifiers for the process detail lines The keys may be Mode Username Imagename Processname Account UIC group Workloadname or PID m DirI O Sec This column provides the average Direct I O rate per second for the process or processes m BufI O Sec This column provides the average Buffered I O rate per second for the process or processes m HrdFit Sec This column provides the average hard pagefault rate per second for the process or processes Img Cnt This column provides the number of image activations as a count for the process or processes m Dsk Ops Sec This column provides the average number of disk I O operations per second for the process or processes Chapter 3 Evaluate Performance in Detail 47 Performance Evaluation Report KB Thruput sec This column provides the average number of Kilobytes transferred to or from disks per second for the process or processes m Top Disk Volume Name IO sec This column provides both the volume name and average I O rate for the disk which this process
280. cs by volume is suppressed PSPASUNNORMALIZE CUSTOM CPU When this logical name is defined to anything custom graphs of CPU utilization depicting more than one node will not have the percentage scaled by the nodes Relative CPU power By default all composite graphs of CPU Utilization are scaled by each node s Relative CPU Power Appendix B Performance Manager Logical Names 409 Appendix C Performance Manager Data Cells Data cells provide access to Performance Manager data for writing analysis rules or for writing your own applications Each data cell entry is displayed in the following format m Data cell name derived m Description m Data Type Domains Target Domain The data cell name is the name used when writing auxiliary rules If a data cell is calculated from other data cells thresholds or provided routines the word derived will be displayed in parentheses after its name The description explains the contents of the data cell The data type describes the format of the data Valid data types are m INDEX SPECIFIER NUMERIC m SCAN ROUTINE STRING m TALLY m TIME Domains identify the valid domains by name from which a rule may reference a data cell When writing rules specify a domain name as the rule s domain to access the data cell This section contains the following topics Data Cell Navigation Table see page 412 Performance Manager Data Cells see page 413 Appendix C Per
281. cs e 50 SCS message and transfer rates e 52 tape activity headers e 58 headings e 58 Granularity e 317 Graph generation modifying weekly schedule e 191 Graph type Buffered I O Images e 191 Compute Queue e 191 CPU Modes e 191 CPU UTILIZATION e 191 Custom e 191 File Cache Hit and Miss Rates e 191 Most Resident Images e 191 Users e 191 Workloads e 191 Number of Jobs e 191 Processes e 191 Physical Memory Utilization e 191 predefined e 191 580 Performance Manager Administrator Guide Systemwide DECnet Traffic e 191 Disk I O Rates e 191 Distributed Locking Activity e 191 Pagefaulting e 191 Terminal I O Rates e 191 Terminal response time e 191 Top BDT Wait Rate e 191 Blk Transfers Requested e 191 BIk Transfers Sent e 191 Buffered I O Users e 191 Buffered I O Workloads e 191 Busy Disk Devices e 191 Busy Disk Volume e 191 Busy Physical Processor e 191 Cluster Rule Occurrences e 191 Compat Mode Procesor e 191 CPU Images e 191 CPU Rule Occurrences e 191 CPU Users e 191 CPU Workloads e 191 Credit Wait Rate e 191 Datagrams Discarded e 191 Datagrams Received e 191 Datagrams Sent e 191 Direct I O Images e 191 Direct I O Users e 191 Direct I O Workloads e 191 Exec Mode Procesor e 191 Faulting Images e 191 Faulting Users e 191 Faulting Workloads e 191 Freeblks Disk Device e 191 Freeblks Disk Volume e 191 Hard Faulting Images e 191 Hard Faulting Users e 191 Hard Faulting Workloads e 191 HSC Channel I O e 191 HS
282. cters Workload Specification The Workloads Excluded list box contains a list of workload definitions To add a workload definition to a workload family click the workload name and click the Include transfer button right arrow To indicate a position within this included workload list click an existing entry in the Workloads Included list box All new workload definitions will be placed ahead of this entry To deselect a position entry click it again All new entries will be placed at the end The position of a workload within a family can determine which workload will include a transaction When a transaction qualifies for more that one workload it will be included in the first listed matching workload To add multiple workload definitions click all their names contained in the Excluded list box Then click the Include transfer button The Workloads Included list box contains a list of workload definitions in the new workload family To remove an entry from the INCLUDED list box click the name and click the Exclude transfer button left arrow To delete multiple workload definitions click all their names and then click the Exclude transfer button Control Buttons To remove all entries and cancel the definition click the Clear button To add a family definition to the defined list and clear the entries click the Create button Modify or Delete a Workload Family To modify a workload family 1 Click a family name The fami
283. ction is produced 6 times summarizing successive 10 minute periods The Interval reports all sections are available from the DCL command line interface Command mode does not provide access to the Interval reports The DECwindows Motif interface allows viewing of interval data with the exception of the Process Section Examples of final tabular reports are shown in the following sections To display the Tabular Report specify the ADVISE PERFORMANCE REPORT TABULAR command Chapter 3 Evaluate Performance in Detail 71 Tabular Report Sections System Configuration Data The following example shows a tabular report with system configuration data Tabular Report YQUEM VAX 6000 440 PA Vx x Tuesday 26 JAN 2006 09 00 to 10 00 Node YQUEM Data collection started 26 JAN 2006 09 00 00 00 Data collection ended 26 JAN 2006 09 02 00 00 Sample interval 120 seconds 2 0 minutes Report generated 12 FEB 2006 17 08 14 35 Processor type is VAX 6000 440 Running VMS version V5 4 3 Total memory 524288 pages 256 00 MB Non paged memory 70012 pages 13 4 5 of total memory Paged memory 454276 pages 86 6 of total memory System working set 16384 pages 3 6 of paged memory User memory paged system working set 437892 pages 96 4 of paged memory 83 5 of total memory System configuration data consists of the following items Node The name of the node for which the system configu
284. ctory and are named SUPPLY_CPU_UTILIZATION REG and SUPPLY_TOP_BUSY_VOLUMES REG ADVISE PERFORMANCE GRAPH TYPE CPU UTILIZATION TOP BUSY VOLUMES FORMAT REGIS CHARACTERISTIC COLOR NODE SUPPLY OUT You can specify device directory and filenames with the OUTPUT qualifier if you do not want to take the defaults Components of Graphs Each graph has the same basic components as shown in the following table Component Applies to Title graphs and pie charts Subtitle graphs and pie charts Axis labels graphs only Chapter 4 Generate Historical Graphs 111 Components of Graphs X and Y axis markers graphs only Legend graphs only Units and Unit total pie charts only MIN MAX and AVG graphs only if one metric or if items are stacked The following list describes each component Title For the predefined graphs the title identifies the type of graph and is centered at the top of the graph For custom graphs the title is PSPA CUSTOM GRAPH unless you specify a title Titles for ReGIS and PostScript graphs are in enlarged characters Subtitle The graph and pie chart subtitle gives the node name or the list of node names for composite graphs and the date and time of the selected data Also the x axis data points and the width in time of each point is provided for graphs Axis Labels All graphs have Time implied as the x axis label Labels for the y axis specify the units of the plotted values for example
285. current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PAGE_CONVERT Derived This contains a one if the hardware model is a VAX 2 if not necessary to scale rules which depend on CPU specific page counts for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP PAGE_WAITMET_F_SPMPAGEWAIT Percentage of time that the CPU was idle and at least one disk device had paging I O in progress for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 475 Performance Manager Data Cells PERCENT CPU TIME IN FILE SYSTEMMET F FILECPU Average percentage of CPU time spent in the file system on the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PGLET CONVERT Derived This contains a one if the hardware model is a VAX 16 if not necessary to adjust rules which mix page and pagelet parameters for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP PORT KB MAPPED Derived This contains the value representing the number of kilobytes mapped per second from t
286. current version of the Performance Manager module Format ADVISE PERFORMANCE SHOW VERSION Example ADVISE PERFORMANCE SHOW VERSION Return Performance Manager version Vx x yymm built dd MMM yyyy The ADVISE PERFORMANCE SHOW VERSION command in this example displays a version of x x yymm 238 Performance Manager Administrator Guide Chapter 7 Use Command Mode Commands This is a reference chapter for the Performance Manager command mode syntax Command mode allows you to specify an analysis period that you may want to investigate and then to interactively view graphs and reports To start a command mode session enter the DCL command ADVISE PERFORMANCE This section contains the following topics ADVISE PERFORMANCE see page 239 SELECT see page 240 LOAD see page 245 GRAPH see page 245 PIE CHART see page 248 REPORT see page 249 SAVE see page 255 SPAWN see page 255 EXIT see page 256 Execute Procedure see page 256 ADVISE PERFORMANCE Format Description The ADVISE PERFORMANCE command invokes a Performance Manager command mode session ADVISE PERFORMANCE When you invoke command mode you see the PSPA prompt At this prompt you can enter the commands listed in the following table You can end a command mode session with the EXIT command Command Function SELECT Causes data to be selected for subsequent viewing by GRAPH and REPORT commands Chapter 7 Use Com
287. d These key levels include IMAGENAME MODE USERNAME UIC_GROUP PROCESS_NAME WORKLOAD_NAME ACCOUNT_NAME and PID For a description of these key levels see the table of focus types in ADVISE PERFORMANCE REPORT The Tabular process metrics require ALL key levels When specifying PID or PROCESS_NAME key levels additional virtual memory may be required See Appendix D Estimating Virtual Memory Needs for more information By default the Performance Manager provides Performance Evaluation processing without Process key levels PID or Processname GRAPHS NO suboption Suboptions include IMAGENAMES USERNAMES HOTFILES USERVOLUMES IO DEVICES BY NODE ALL and DEFAULT DEFAULT consists of IMAGENAMES USERNAMES and IO DEVICES By default the Performance Manager provides GRAPH DEFAULT processing Qualifiers AVERAGE DAILY WEEKLY MONTHLY QUARTERLY Causes graphs to depict a specified time period The selected data is averaged into the time period selected If you also use the SCHEDULE qualifier the DAILY and WEEKLY graphs are trimmed to show only the selected hours The DAILY and WEEKLY graphs must select data from at least two different days and the MONTHLY and QUARTERLY graphs must select data from at least two different months If history data with the periodicity attribute is selected the AVERAGE value is automatically set to that periodicity value This is true regardless of whether the AVER
288. d as a rate per hour The metric graphed is IO OCCURRENCES TYPE TOP IO VOLUMES Top Operations Disk Volume Plots the five plus Other Volumes disk volumes that incurred the highest I O rates TYPE TOP KB MAP Top Kilobyte Mapped Rate Plots five nodes plus Other in terms of the number of kilobytes of data mapped for block transfer The metric graphed is KB MAP TYPE TOP KB RC Top Kilobyte Received Rate Plots five nodes plus Other in terms of the number of kilobytes of data received by the local system from the remote system through request data commands The metric graphed is KB RC TYPE TOP KB S Top KB Sent Rate Plots five nodes plus Other in terms of the number of kilobytes of data sent from the local system to the remote system through send data commands The metric graphed is KB S TYPE TOP KERNEL PROCESSOR Top Kernel Mode Processor Plots the five plus Other processors in terms of time spent in kernel mode as a percent of CPU time The metric graphed is P KERNEL TYPE TOP MEMORY RULE OCC Top Memory Rule Occurrences Plots the five plus Other memory rule identifiers that fired as a rate per hour The metric graphed is MEMORY OCCURRENCES and is available only from history data 224 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE TOP MGS R Top Messages Received Plots five nodes plus Other in terms of number of application da
289. d choose the Assign Metrics menu item Performance Manager displays an Instrument Metric Selections dialog as shown in the following screen User DECNET MAIL EX File Commands Edit Options Help CPU Utilization Class Name PASA Single CPU Utilization Disk IO Rate Disk TP W riter Metric Hard Page Faults Soft Page Faults Unsorted e O include Class in Label Metric Alias O use Alias in Label Ki When Assign Metrics is chosen the first metric field in the instrument to be assigned a metric will appear with a solid outline If the instrument has more than one field the others will appear with a dashed outline Chapter 10 Customize the DECwindows Motif Real time Display 359 How You Edit the Panel Instruments The Instrument Metric Selection dialog lets you perform the following actions m Select a metric class m Select a metric name m Sort metric classes in ascending or descending order m Filter metric selections by value or instance m Include the class name in the instrument label m Specify a metric alias a user supplied string which is displayed instead of the metric name The first metric field will automatically be selected as shown by the solid outline If another field is desired click the instrument s outlined metric field You may have to move the dialog out of the way if it occludes the instrument being modified The Instrument Metric Selections dialog displays the instr
290. d faulting e 191 throughput e 191 interpret process statistics e 38 largest VA space e 191 WS size e 191 most image activations e 191 resident e 191 used with performance evaluation reports e 33 Workload family definitions add workloads e 314 create e 314 define e 313 modify e 313 order precedence e 314 remove workloads e 314 workloads included e 314 WritIN Prog e 79 X X POINTS keyword e 191 XQP e 81 Y Y AXIS MAXIMUM keyword e 191 245 YESTERDAY keyword e 191 Index 591
291. d page list before the swapper writes this list to acquire free pages for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MPW WAITLIMITPAR F MPW WAITLIMIT The value of the SYSGEN parameter MPW WAITLIMIT which sets the number of pages on the modified page list that will cause a process to wait until the next time the modified page writer writes the modified list for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MPW WRTCLUSTERPAR F MPW WRTCLUSTER The value of the SYSGEN parameter MPW WRTCLUSTER which sets the number of pages to be written during one I O operation from the modified page list to the page file or a section file for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MULTIPROCESSINGPAR F MULTIPROC The value of the SYSGEN parameter MULTIPROCESSING which enables full checking synchronization for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC
292. d process device IO rates For the Motif displays you can view several default panels that refer to this additional information see Chapter 10 for these panels with MODE DISKFILE Chapter 6 Performance Manager Commands 179 ADVISE PERFORMANCE DISPLAY INTERVAL seconds C W Specifies the interval in seconds between collection records when collecting data in NETWORK mode The INTERVAL qualifier is ignored when you specify MODE DISKFILE the interval is that of the collection selected The valid range is 1 to 86400 RULES file spec C With the character cell interface generates brief conclusions based on a knowledge base you have specified to supplement the default knowledge base The file spec must point to the auxiliary knowledge base which has previously been compiled with the ADVISE PERFORMANCE COMPILE command The default file type is KB If you specify NORULES no auxiliary knowledge base is used for the display even if AUTO AUGMENT is enabled VOLUMES name C The VOLUMES qualifier allows you to specify a list of disk volume names that is included in the Real time Character Cell displays If the qualifier is omitted all disk devices are included on the displays Server statistics are provided for the servers that provide access to the disks selected This qualifier is mutually exclusive with DISK DEVICES INITIAL options C With the character cell interface indicates which display to bring u
293. data collector in a swapping file resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NUMBER OF INSWAPPED PROCESSES Derived This contains a value representing the number of processes in the balance set on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NUMBER OF OUTSWAPPED PROCESSES Derived This contains a value representing the number of processes not in the balance set on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NUMBER OF PROCESSESMET F PROCCNT This contains a value representing the average number of processes on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP NUM_PROCS_NOT_USING_WS_LOANS Derived This contains the value indicating a count of processes not using working set loans on the local node for the current interval Data Type NUMERIC 472 Performance Manager Administrator Guide Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP OPEN FILESMET F OPEN FILES Average number of open files for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP OUTGOING_BLOCKING_AST_RATEMET_F_BLK_OUT This contains the number of outgoing blocking ASTs queued per second for the current interval for the loc
294. data on standalone systems or clusters running OpenVMS Performance Manager analyzes statistics and parameters collected by Performance Manager from each node in a configuration to determine whether specific conditions are contributing to system performance degradation Based on its findings Performance Manager recommends ways to improve system performance and provides evidence to support its conclusions Performance Manager can organize your information into several different reports that can do the following actions m Identify system resource limitations when they exist for the workload m Identify system parameter settings that may be adding to system overhead or degrading system performance m Evaluate trends in system performance m Evaluate the effects of changes in workload and configuration characteristics 14 Performance Manager Administrator Guide Performance Manager Features Performance Manager Features Performance Manager has the following major features m Performance knowledge base and rules compiler m Performance analysis and reports m Real time displays of performance data Graphing m Data export facility Knowledge Base and Rules Compiler The Performance Manager knowledge base consists of rules and thresholds used to evaluate system performance The rules provided at installation time are known as factory rules When Performance Manager produces an Analysis Report by default it uses the factory rules Per
295. ded Process Disk or Server Statistics sections for node reporting Cluster wide reports are also provided for summary process disk and server reports which focus on either a cluster view or a by node view The window is sized to accommodate the requirements of the Summary Statistics display The section label changes according to what section is currently viewed as are the arrow buttons enabled or disabled Pressing the node button labeled LATOUR in the figure displays an option menu with the list of selected node names for which the report can be viewed Releasing on a node name causes the report window to be updated with that node s data The arrow buttons on either side of the node button can be used to progress sequentially through the selected nodes 284 Performance Manager Administrator Guide How You Display Analyzed Data The node option menu and arrow buttons are not enabled while cluster reports are being viewed The process and disk report sections can generate many screens worth of data When viewing these report sections the number of screens available for viewing and the arrow buttons on either side become enabled to allow reviewing all available data To exit from the Tabular Report Window pull down the File menu and release on Close Tabular Interval Report Sections In addition to viewing the tabular report statistics summarized over the selected analysis period you can also view classes of statistics according to
296. dicates that the extended process metric data is to be exported along with the standard process metric data If you do not specify EXTENDED then NOEXTENDED is assumed The presence of the IMAGE keyword indicates that the image name for each process is to be exported along with the standard process metric data If you do not specify IMAGE then NOIMAGE is assumed The presence of the ALL keyword indicates that all available process metric data is to be exported including the standard process metric data the extended process metric data and the image name If you omit the CLASS keyword from the command line then CLASS DEFAULT STATISTICS is assumed CLASSIFY BY USERGROUP workload family Lets you specify a workload family to control how process data is classified By default all process data is exported without being summarized Chapter 6 Performance Manager Commands 183 ADVISE PERFORMANCE EXPORT COLLECTION DEFINITION collection definition name Specifies the name of the Collection Definition and hence the collected data that you want to export If you omit this qualifier daily data is obtained from the Collection Definition called CPD To view the Collection Definitions that you have available use the DCL command ADVISE COLLECT SHOW SCHEDULE If you want to use history data instead of daily data use the HISTORY qualifier instead of the COLLECTION DEFINITION qualifier These two qualifiers are mutually exclusive DATES
297. discover and resolve performance problems Although you need the information Performance Manager provides for investigating any perceived performance problem this information is not infallible Occasionally Performance Manager recommendations do not improve performance Additional expertise analysis hardware and tuning may be required to solve a specific performance problem Cross Platform Support If you run Performance Analysis across platforms between r3 1 and a VAX or Alpha system running prior releases of Performance Agent you need to be aware of the following situations m CA does not support using an older Performance Manager r3 or older with the new Performance Agent r3 1 m The Performance Manager r3 1 runs performance analysis on Alpha and VAX data from prior releases However the results might be different from those generated on prior releases due to the updated and new rules m The VAX Performance Manager does not analyze r3 1 data Chapter 1 Introduction 19 Chapter 2 Analyze Performance This chapter contains example Performance Manager Analysis Reports and information to help you interpret them The chapter Performance Manager Commands see page 175 explains how to generate these reports This section contains the following topics Analysis Reports see page 21 Brief Analysis Reports see page 30 Analysis Reports When you request an Analysis Report the Performance Manager analyzes data for the
298. ds their qualifiers keywords and options See the Performance Agent Administrator Guide for a complete description of the ADVISE ARCHIVE ADVISE COLLECT and ADVISE EDIT commands At installation time the Performance Manager software adds the ADVISE PERFORMANCE command to the DCL command table To start a Performance Manager action issue the ADVISE PERFORMANCE command with the appropriate action option to perform the desired task for example ADVISE PERFORMANCE COMPILE This section contains the following topics ADVISE PERFORMANCE see page 175 ADVISE PERFORMANCE COMPILE see page 176 ADVISE PERFORMANCE DISPLAY see page 178 ADVISE PERFORMANCE EXPORT see page 182 ADVISE PERFORMANCE GRAPH see page 191 ADVISE PERFORMANCE PIE CHART see page 232 ADVISE PERFORMANCE REPORT see page 234 ADVISE PERFORMANCE SHOW VERSION see page 238 ADVISE PERFORMANCE Format Description The ADVISE PERFORMANCE command initiates the functions of the Performance Manager module ADVISE PERFORMANCE option ADVISE INTERFACE DECWINDOWS MOTIF The ADVISE PERFORMANCE command options are described individually in this guide If you do not specify an option the command defaults to command mode with a PSPA prompt For more information about command mode see the chapters Customize the Knowledge Base see page 135 and Use Command Mode Commands see page 239 Chapter 6 Performance Manager Commands 175
299. ds One Line Description 22 Low hit ratio high attempt rate on the file header cache 0 1 Lots of contention for distributed locks Summary for SUPPLY 2 rules fired of 338 records 23 satisfied conditions Summary for VOLTY O rules fired of 337 records 1 satisfied conditions CLUSTER ID oftime Recds One Line Description 1 1 0 bottleneck on disk reduce or redistribute load Summary for CLUSTER 1 Rules fired Chapter 2 Analyze Performance 31 Chapter 3 Evaluate Performance in Detail This chapter contains example Performance Manager statistical reports and information to help you interpret them For more information about obtaining Performance Manager reports see the chapter Performance Manager Commands see page 175 This section contains the following topics Performance Evaluation Report see page 33 Histograms see page 64 Tabular Report Sections see page 71 Performance Evaluation Report The Performance Evaluation Report provides statistics on system use component use and process activity It also provides metrics for performance improvement or degradation to use when evaluating the impact of recommendations made by the Performance Manager The Performance Evaluation Report has the following sections m Process statistics by primary and secondary keys m Pool statistics m CPU mode statistics m SCS statistics m Lock statistics m CI NI and adapter statistics m Disk statistics
300. e For more information see the SET AUTO AUGMENT command described in the Performance Agent Administrator Guide Version Limit Enter a decimal number to modify the file version limit on the Performance Manager parameters file and history files The initial setting is 180 View the Main Window To view the main window Pull down the View menu and choose the Main Window sections you want to display Depending on your processing mode not all the sections in the Main window may be relevant or of interest To allow for smaller windows and the elimination of distracting sections the View menu contains a series of toggle buttons that can be set or reset to add or remove sections of the Main window Resetting a toggle button causes a section to disappear setting it restores the section Do not remove the File Locks section if you share editing access of database files with other users 320 Performance Manager Administrator Guide Chapter 9 Use the DECwindows Motif Real time Display This chapter provides information about the use and basic modification of the default displays supplied with the Performance Manager real time display These predefined displays consist of windows or instrument panels containing bar graphs strip charts or meters These instruments are used to view OpenVMS system performance statistics To access performance data for use with the Real time Display you must have the Performance Manager installed and a
301. e Batch Network or Detached Process statistics detail lines by the process s User name Process statistics detail lines by the process s Image name Chapter 7 Use Command Mode Commands 253 REPORT FOCUS Types Provide UIC GROUP Process statistics detail lines by the process s UIC Group PROCESS NAME Process statistics detail lines by the process name WORKLOAD NAME Process statistics detail lines by the workload name You must specify CLASSIFY BY to indicate the workload family that you intend to use ACCOUNT NAME Process statistics detail lines by the process s account name PID Process statistics detail lines by the process s PID Other Options Provide NO CLUSTER A summary of process data for the entire cluster scaled by CPU speed The default is NOCLUSTER NO BY NODE Per node detail of cluster process data The default is NOBY NODE if you specify CLUSTER If the value for the secondary key is the same as the primary key no secondary level breakout occurs This also happens if you specify the primary key and no secondary key is given The CLUSTER and BY NODE keywords allow you to specify that the process statistics section of the Performance Evaluation Report is to present data combining process information for all selected nodes CLUSTER and if so whether the BY NODE detail should also be included By default process data is not combined for all selected nodes When the CLUSTER option is used
302. e and the HSC s channel number for example HSCOO1 6 The following tables identify the custom graphing metrics grouped by metric class Channel CHANNEL IO CHANNEL QUELEN CHANNEL THRUPUT CPU P BUSY Description Number of I O operations transferred by the HSC K SDI channel Number of I O operations outstanding to all disks on the HSC K SDI channel Number of bytes per second transferred by the HSC K SDI channel Description Percentage of time that the physical CPU was busy Chapter 6 Performance Manager Commands 205 ADVISE PERFORMANCE GRAPH Channel P COMPAT P EXEC P IDLE P INTERRUPT Description Percentage of time that the physical CPU was in compatibility mode Percentage of time that the physical CPU was in exec mode Percentage of time that the physical CPU was idle Percentage of time that the physical CPU was in interrupt stack mode P KERNEL Percentage of time that the physical CPU was in kernel mode P MP SYNCH Percentage of time that the physical CPU was in MP synch mode P SUPER Percentage of time that the physical CPU was in supervisor mode P USER Percentage of time that the physical CPU was in USER mode Disk Description BUSY Percent of time that there was one or more outstanding I O operation to the disk D IO SIZE Number of 512 byte pages per I O D RESPONSE TIME SPLITIO FREEBLKS MSCPIO PAGIO PRCT FREE PRCT USED 206 Performance Manager Administrat
303. e is defined to anything PSPASGRAPH CHARS A user defined logical which specifies a string of six characters to be used in place of the normal ANSI graph legend characters 406 Performance Manager Administrator Guide PSPASGRAPH FILE DEVICE PSPASGRAPH FILE DEVICE A user defined logical which when defined to anything causes the graph metrics for File Names to be displayed by file name and device PSPASGRAPH FILE DIRECTORY A user defined logical which when defined to anything causes the graph metrics for File Names to be displayed by file name and directory PSPASGRAPH LEGEND FONT POINT Define this logical name to a number specifying a font point size for the PostScript graph item labels If the PostScript Graph labels are longer than 24 characters and you do not want the labels truncated use this logical name to cause the labels to not truncate and to be reduced in size so as to fit on the page By default the graph legend font point is 10 Defining PSPASGRAPH LEGEND FONT POINT to 7 should reduce the font sufficiently and a 5 makes the font size very small but accommodates up to 50 characters PSPASGRAPH PATH PSPASHLS When this logical name is defined to anything the graph metrics for SCS Nodes is displayed by pathname instead of nodename This provides more detailed information about the load on each adapter for a given node Specifies ReGIS HLS encodings enabling user specifications of color planes for the Perf
304. e 191 CLASSIFY BY e 191 COLLECTION DEFINITION e 178 191 COMPOSITE e 191 DATES e 182 191 DISK DEVICES qualifier to create displays e 178 DNS NAMES qualifier to create displays e 178 ENDING e 191 FILTER to select data in command mode e 241 FILTER qualifier for pie charts e 232 to export data e 182 to graph data e 191 FORMAT e 117 191 HISTORY e 191 INITIAL qualifier for rule conclusions e 178 INTERVAL qualifier to creat displays e 178 LINE e 191 MODE qualifier for creating displays e 178 NODE e 191 NODE NAMES qualifier for generating graphs e 191 to create displays e 178 OUTPUT e 191 OUTPUT qualifier to generate graphs e 191 PERCENTAGE e 232 249 PROCESS STATISTICS e 34 to export data e 182 RULES qualifier Index 585 for knowledge base file name e 177 for rule conclusions e 178 SCHEDULE e 191 TYPE e 108 191 245 to generate all graphs e 245 TYPE qualifier to generate all graphs e 191 VOLUMES qualifier for rule conclusions e 178 Queue length analysis report e 29 disk devices e 191 disk volumes e 191 HSC channels e 191 R Ranges and thresholds set o 338 347 Real time character cell displays start e 374 displays character cell e 178 windows e 178 panel manager display of e 341 Playback display from DECwindows e 322 sessions terminate e 345 Recommendation e 18 236 from analysis report e 21 Recovery procedures e 403 ReGIS graph e 117 keyword e 117 191 Remove panel instruments e 342 Rename
305. e 384 Display Rules Information see page 385 Display RESOURCE Information see page 385 The INVESTIGATE Command see page 391 Evaluate Performance Using the Investigate Displays see page 392 Exit the Character Cell Displays see page 401 Character Cell Display Functions Prerequisites The Performance Manager character cell displays gather and present performance data using a video terminal Some displays are available on terminals that support DEC CRT characteristics such as the VT100 Use the SET TERMINAL DEC CRT command to set characteristics for these terminals Other displays are available only for ReGIS compatible terminals such as the VT340 If the terminal supports color or if an external color monitor is attached a multicolored display is generated The display can also be printed on a graphics dot matrix printer The Performance Manager character cell displays have the following mandatory software and hardware requirements m For real time remote data collection MODE NETWORK command see the discussion about establishing remote access in the Performance Agent Administrator Guide m For file access MODE DISKFILE command SYSLCK privilege is required Chapter 11 Use the Character Cell Real time Display 373 Start the Character Cell Displays m A ReGIS compatible terminal such as the VT125 VT240 VT241 VT330 or VT340 is needed for most displays invoked through the INVEST
306. e By Value option To enable filtering based on a specific data item identifier such as a specific user name process name disk name and so on select the By Instance option Chapter 10 Customize the DECwindows Motif Real time Display 361 How You Edit the Panel Instruments Assign Launch Panels For the By Value option you can then specify the range of values to be displayed using the Filter option menu Press and hold MB1 over the Filter options to see the following range of options m Less m Less or Equal m Equal m Not Equal m Greater or Equal m Greater m In Range m Out of Range The value to be compared for range determination can be entered in the Compare Value fields or can be based on a value from a parent panel as described below If the Filter metric is a string then the value specified is interpreted as all uppercase unless the string is contained within double quotes If the instrument being modified is launched from another panel then a value may be implicitly passed to this panel In this case this value is considered the instance value See the Launching Panels section for more information If you want to use this instance value select By Instance in the Type option menu and select the applicable Filter Metric Name corresponding to the passed value and leave the Compare Value field empty Clicking on the Apply button applies your choices without closing the dialog The OK button applies your choices a
307. e HSC Description Number of I O s per second to this file Number of paging I Os per second to this file Number of read I O s per second to this file Chapter 6 Performance Manager Commands 207 ADVISE PERFORMANCE GRAPH Channel FILE WRITIO FILE THRUPUT FILE RESPONSE TIME Description Number of write I O s per second to this file Number of bytes per second transferred to or from this file Average number of milliseconds elapsed between the start of the IO SIO and its completion EIO for all of the I Os to the file FILE SPLITIO Number of split I O s per second to this file Process Description BUFIO Number of process buffered I O operations per second CPUTIME Percent of total CPU time that the process es consumed DIRIO Number of process direct I O operations per second DSKIO Number of process disk I O operations per second DSKTP Number of process bytes per second transferred to and from disks FAULTS Number of process hard and soft page faults per second HARDFAULTS Number of process page fault I O IMAGE ACTIVATIONS IO SIZE RESIDENCE RESPONSE TIME 208 Performance Manager Administrator Guide operations per second Number of process image activations per second Average number of pages per process disk I O Number of resident processes with either the specified user name or image name Average number of seconds between the end transaction for a terminal read
308. e Manager uses the alphabetic characters in the following table as the first letter of the rule identifier The following table lists the rule ID abbreviations Letter Rule Performance Category C CPU related rule M Memory related rule I I O related rule R Resource miscellaneous rule L Cluster related rule X XFC related rule Chapter 5 Customize the Knowledge Base 149 Components of Rules Domain Element The domain element defines the context in which the rule exists The association between a rule and its domain designates when a rule is tested and under which Analysis Report section the rule firing is reported The definition of a domain for Performance Manager knowledge base processing is based on the sub records that are read from collected data Each rule exists within a single domain Also each data cell is associated with a set of domains Rules can reference any data cells in the domain in which the rule exists If you omit the domain element the default domain for the rule is LOCAL The format of the domain element is as follows DOMAIN domain name TRACE domain name Can be one of the following names m CLUSTER m COMMUNICATION m CONFIGURATION m CPU m DISK m FILE m LOCAL m LOCALC m PROCESS m SUMMARY m TAPE Domain Names Rule Testing and Reporting Domain selection determines the data against which the rule is evaluated the frequency with which the rule is tested and the Analysis Report sect
309. e a copy of each in your private directory to modify do not edit these files directly Disable an Existing Rule Assume that you work in a secure government facility and data security is a top priority You do not care if security erase I Os exceed their default threshold of 1 I O per second and would like to disable the rules that check for this To do this first check in the source file of the factory rules PSPAS EXAMPLES PSPA KB VPR and in the message text file SYSSSYSTEM PSPASMSG TXT for the rules of interest You find one rule 10180 which focuses on security erasures To eliminate this rule from further consideration in your knowledge base add the following line to your rules source file MYRULES VPR DISABLE I0180 The semicolon terminator is important if you forget one of these your auxiliary rules file probably does not compile To use this rule file see the sections Build an Auxiliary Knowledge Base see page 172 and Use an Auxiliary Knowledge Base for Reporting and Archiving see page 173 Modify an Existing Rule Assume that you do not want to see all the lines of evidence produced by rule R0095 for the image VMSBUXX because you cannot change the operation of this image To modify a rule you need to first disable the factory version of that rule then copy the factory version into your MYRULES VPR file and modify it accordingly Make sure that you define all literals used in this rule see the secti
310. e a units label for the bars Specify the space for the units label in characters Label Fonts The Font Selection dialog lets you specify label font characteristics Location of Tags and Ticks Specifies whether tags numeric values indicating a chart s scale and tick marks are displayed on the left or right side or top or bottom of the chart Tick marks can be displayed without tags When tags are selected tick marks are displayed automatically 354 Performance Manager Administrator Guide Strip Chart Parts How You Edit the Panel Instruments Number of Tags Ticks Specifies the number of tags and corresponding tick marks to be disp Numbe layed The default is six tags r of Short Ticks Specifies number of minor tick marks to be displayed between the major tick Tit i BE Data Name Tags Tick Marks s u Minutes BE Bottom Side INE Tag Sho Units marks set in Number of Tags Ticks The default is one Left Right Display Tick Marks a B Hours IN Top Side J Seconds CJ s Tick Marks 6 rt Tick Marks 1 Label Units Label CJ The Stri Title Set p Chart Parts dialog lets you specify the following attributes the toggle button to display a title Enter a title in the text entry box Data Name Set any the toggle button to specify the title to be the metric name overriding supplied string Title Font The Font Selection dialog lets you specify the f
311. e discussed next RULE Constructs define rules The format of the rule construct is as follows RULE rule elements ENDRULE The rule construct consists of seven rule elements The optional elements in the list following are enclosed in brackets Rule ID element Domain element Rule condition element Rule condition element Occurrence threshold element Evidence element Conclusion text element Brief conclusion text element Once defined and compiled the rules can be supplied to the analysis process to cause a conclusion to be fired as appropriate or can be supplied to the archival process to cause the number of occurrences of the rule to be archived DISABLE Constructs allow you to turn off individual rules The format of the disable construct is as follows DISABLE rule id rule id The parameter rule id specifies which rule or rules to disable the construct ends with a semicolon LITERAL Constructs allow you to define unique symbols for use in a rule expansion The format of the literal construct is as follows LITERAL literal definition literal definition ENDLITERAL A literal definition is defined as literal symbol decimal value A literal symbol is a unique symbol that you have equated to a decimal value A literal symbol can contain up to 40 alphanumeric characters that have not been defined before as a literal a threshold or a data cell A literal symbol can be
312. e for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK INTERVAL MSDEV F ITVL This contains a value representing the uptime of the disk in milliseconds for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK 436 Performance Manager Administrator Guide Performance Manager Data Cells DISK IO RATEDEV F OPCNT This contains a value representing the average number of I O requests per second to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK IO RATE TALLY Derived This contains the sum of the values representing the average I O rate for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_IO_RATE_THRESHOLD Derived This contains a value representing the disk I O rate threshold for the current disk during the current interval This value is obtained from the threshold TD_Tn_xxxx where n is the disk type model number found in STARLET DCDEF for the current disk and xxxx is its type e g TD_T21_RA81 Data Type NUMERIC Domains CLU DISK_IS_SERVED Derived This contains a Boolean value indicating whether the current disk is MSCP served during the current interval Data Type NUMERIC Domains CLU DISK_MAP_CACHE_SIZEDEV_F_MAPSIZE Thi
313. e for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SRPS_IN_LISTMET_F_SRP_MAX This contains the total number of SRPs for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 500 Performance Manager Administrator Guide Performance Manager Data Cells SRPS IN USEMET F SRP CNT This contains the number of SRPs in use for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SRP EXPANSION COUNT Derived A count of the number of times that the number of small request packets needed to be increased for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY SRP MAXLEN Derived The maximum size of the SRP list for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY STORAGE MAP CACHE AR Derived Attempt rate per second to the storage bit map cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP STORAGE_MAP_CACHE_HR Derived Hit ratio to the storage bit map cache for the local node for the current interval record Calculated by dividing the number of storage bit map cache hits by the number of storage bit map cache attempts hits misses times 100 Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SUPERMET_F_SUPER Average percentage of CPU
314. e for the display for doing playback Normally you do not need to specify this qualifier however to view previously recorded data using the Real time displays specify the desired begin time When completed with the display use the DISCONNECT and EXIT push buttons to exit You can specify either an absolute time or a combination of absolute and delta times For complete information on specifying time values see HP s OpenVMS User s Manual or type HELP DATE TIME You can also use the keywords TODAY TOMORROW and YESTERDAY BEGINNING is mutually exclusive or ignored with MODE NETWORK 178 Performance Manager Administrator Guide ADVISE PERFORMANCE DISPLAY COLLECTION DEFINITION collection definition name C W Specifies the name of the Collection Definition and hence the collected data that you use for the dynamic display If you omit this qualifier data is obtained from the Collection Definition called CPD Use the ADVISE COLLECT SHOW STATUS command to see which collection definitions are active The COLLECTION DEFINITION qualifier is used in conjunction with the MODE DISKFILE qualifier and is ignored if you specify MODE NETWORK DISK_DEVICES devicename C W The DISK_DEVICES qualifier allows you to specify a list of disk device names that are included in the Real time Character Cell displays If the qualifier is omitted all disk devices are included on the displays Server statistics are provided for t
315. e of XYZZY If the value is greater than O the rule expression is true Chapter 5 Customize the Knowledge Base 159 Data Cell Types and Use These scan routine data cells are valid COMMUNICATION SCAN m CONFIGURATION SCAN m CPU SCAN m DISK SCAN m FILE SCAN m PROCESS SCAN m TAPE SCAN You cannot use scan routine data cells as evidence Tally Data Cell Tally data cells contain data tallied from those sub records scanned when the scan routine rule expression holds true Each scan routine has a target domain Each tally data cell has a target domain A given scan routine updates those tally data cells with the same target domain as the scan routine s target domain Scan routine data cell names start with the target domain name For a given rule you can use tally data cells in a rule expression or evidence after a call to a scan routine For example a LOCAL domain rule with the following expression makes available the process TALLY fields that contain the sum of all PROCESS domain metrics for process sub records with the user name of CHARLIE PROCESS SCAN USER NAME EQS CHARLIE GT 0 Subsequent rule expressions may use the tally data cells in the same rule For example PROCESS CPUTIME TALLY GT 1000 Tally Data Cells and Associated Scan Routines The following table shows tally data cells and the associated scan routines Scan Routines Tally Data Cells Updated by Scan Routines COMMUNICATION SCAN COMM OPERATI
316. e of a large process in computable state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LARGE NOSWAP PROCESS EXISTS Derived This contains a Boolean value representing the presence of a large process where the PSWAPM privilege inhibited swapping on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LARGE NOSWAP PROCESS X Derived This contains an index pointing to the process subrecord for a unique username for a large process with the PSWAPM privilege enabled on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO LARGE_PROCESSES_ EXIST Derived This contains a Boolean value representing the presence of one or more large processes on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 463 Performance Manager Data Cells LCK EXPANSION COUNT Derived A count of the number of times the LOCKIDTBL needed to be extended when the system ran out of LOCKIDTBL entries for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY LCK MAXLEN Derived The maximum number of entries in the Lock ID table for the local node for all of the intervals Data Type NUMERIC Domains SUMMARY LOCAL BLOCKING AST RATEMET F BLK LOC This con
317. e or volume name m Processor mode metrics selected by processor ID m HSC metrics selected by HSC node name m SCS metrics selected by SCS cluster node name Rule metrics for archived data only selected by rule ID m HSC channel metrics selected by HSC channel name 126 Performance Manager Administrator Guide Generate Custom Graphs m File metrics selected by file name m Process Disk Volume I O Rates selected by user name and volume name or image name and volume name These metrics are described in the chapter Performance Manager Commands see page 175 You can specify the metrics and data items by one of the following methods m DCL command m Interactive prompting The units that the various metrics represent may differ for example I Os per second or percentage of CPU time The Performance Manager allows you to include data with different metrics on the same graph Use your discretion when doing this however Graph System Metrics The following command generates a custom graph ADVISE PERFORMANCE GRAPH TYPE CUSTOM _ SYSTEM METRICS DZRO GVALID TITLE Pagefaulting The following example Prompting for System Metrics Custom Graph shows Performance Manager prompts and user input that generate the same graph in command mode ADVISE PERFORMANCE PSPA gt SELECT BEGINNING 19 FEB 2006 13 00 _PSPA gt ENDING 19 FEB 2006 17 00 GRAPH BY NODE The resulting graph appears in the example Custom Gra
318. e representing scaling factor for the compute queue length for the local node The value is obtained from the threshold TD_COM_SCALING_n where n is the hardware model number of the local node By default if the local node is a VAX 11 780 the value would be 1 0 The value of this data cell can be modified using a threshold construct Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CONFIGURATION_SCAN Derived Provides the count of configuration subrecords for which the specified rule condition is true The condition will be evaluated for each configuration subrecord Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG Appendix C Performance Manager Data Cells 423 Performance Manager Data Cells CPUIO BUSYMET F SPMCPUIO Percentage of time that both the CPU and at least one disk device were busy for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CPUIO IDLEMET F SPMSYSIDLE Percentage of time that the CPU and all disk devices were idle for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CPU BUSYMET F SPMBUSY Percentage of time that the CPU was busy for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CPU COMPATCPU F COMPAT This contains the value representing the percent of time spent in COMPAT
319. e uptime for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO Appendix C Performance Manager Data Cells 487 Performance Manager Data Cells PROCESS VIRTUAL PAGESPRO F VA USED This contains a value representing the number of virtual pages used by this process This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN CEFPRO V SSS CEF This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the common event flag wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN COLPGPRO V SSS COLPG This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the collided page wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN COMPRO V SSS COM This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the compute queue This cell pertains to the current process subrecord on the
320. ecent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG KERNELMET_F_KERNEL Average percentage of CPU time spent in Kernel mode for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LARGEST_BLK_IN_NONPAGED_POOLMET_F_NP_MAX_BLOCK This contains the number of bytes in the largest block in non paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LARGEST_BLK_IN_PAGED_POOLMET_F_PG_MAX_BLOCK This contains the number of bytes the largest block in paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LARGEST_WS_PROC_X Derived This contains an index pointing to the process subrecord for a unique username with the largest working set on the local node for the current interval Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO 462 Performance Manager Administrator Guide Performance Manager Data Cells LARGE BATCH PROCESSES EXISTS Derived This contains a Boolean value representing the presence of one or more large batch processes on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LARGE COM PROCESS EXISTS Derived This contains a Boolean value representing the presenc
321. ed in parentheses by the processor type or HSC type with which they are associated Node names are followed in parentheses by the processor type for the node The following example shows server statistics in BYNODE format Paging Swping By Cluster and By Node Server Statistics The server statistics reported in By Cluster and By Node formats are Alloc Class The allocation class of the node or HSC associated with the server Reported only for By Cluster statistics Work Avail The percentage of time there were I O requests at the server queue Reported only for By Node statistics Paging The percentage of the work available marked as Page I O Swping The percentage of work available marked as Swap I O Queue Length The average sum of the requests at the server queue 106 Performance Manager Administrator Guide Chapter 4 Generate Historical Graphs You can generate graphs and pie charts from current or historical data for many aspects of the system Numerous types of predefined graphs are available You can also create custom graphs to represent your site specific needs These graphs can be printed or displayed on your terminal The FILTER qualifier lets you select a subset of data for graphs For more information see the chapter Performance Manager Commands see page 175 This section contains the following topics Generate Predefined Graphs see page 107 Generate Multiple Graphs see page 111 C
322. ed pool Number of free non paged pool blocks less than or equal to 32 bytes in size Size in Kbytes of largest free non paged pool block Size in bytes of smallest free non paged pool block Size in Kbytes of non paged pool Number of NV terminal I O operations per second Number of buffered I O operations less any terminal I O operations per second Number of outgoing ENQs or Lock Conversion CVTs to remote nodes per second Number of active outswapped processes COMO Channel OUTPROCINACT PAGEFILE UTILIZATION PFW PG FREE BLOCKS PG FREE BYTES PG FREE LEQ 32 PG MAX BLOCK PG MIN BLOCK PG POOL MAX ADVISE PERFORMANCE GRAPH Description Number of inactive outswapped processes Percent of pagefile pages in use or occupied Average number of processes in page fault wait state Count of paged blocks Number of free Kbytes in paged pool Number of free paged pool blocks less than or equal to 32 bytes in size Size in Kbytes of largest free paged pool block Size in bytes of smallest free paged pool block Size in Kbytes of paged pool PREADIO Read operations per second from a disk due to a page fault PREADS Pages read per second from a disk due to a page fault PWRITES Pages written per second to paging files PWRITEIO Write operations per second to paging files RCVBUFFL Receiver buffer failures per second RELATIVE CPU POWER This node s VUP rating as a percentage o
323. ed by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESSES IN CEFMET F CEF This contains a value representing the average number of processes in common event flag wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES_IN_COLPGMET_F_COLPG This contains a value representing the average number of processes in collided page wait state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 477 Performance Manager Data Cells PROCESSES IN COMMET F COM This contains a value representing the average number of processes in the computable state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES IN COMOMET F COMO This contains a value representing the average number of processes in the outswapped computable state on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES IN CURMET F CUR This contains a value representing the average number of processes in the currently executing state on the local node for the current interval Sampled every 5 seconds
324. ed image records for this process are summarized 4 The first 12 characters of the image name string associated with the user s process 5 Average working set size in pages for the user s process while running the specified image 6 Total number of page faults for the user s process while running the specified image divided by the CPU seconds for the same period 7 Average system wide hard page fault rate during the major sampling intervals when the user s process ran the specified image 8 Average system wide soft page fault rate during the major sampling intervals when the user s process ran the specified image 9 Average size of the free page list in pages during the periods when the user s process ran the specified image 10 Number of times that a Performance Manager process or image record supports the evidence In the previous Memory Rule report example the total number of page faults per CPU second are in the range of 268 to 363 for user Smith and 2046 for user Correy This high rate of page faulting probably contributed to the system wide soft page fault rate ranging from 106 to 791 which exceeded the threshold of 100 This occurred 13 times 4 5 2 2 because Smith s WSEXTENT was too low at the current value of 200 and Correy s WSEXTENT was too low at its value of 4000 24 Performance Manager Administrator Guide Analysis Reports CPU Rule Analysis Report The following example shows a CPU Rule analysis Fu
325. ed to match the largest data value shown Maximum Data Value Specifies the maximum value of the chart s scale The default value is 100 Minimum Data Value Specifies the minimum value of the chart s scale The default value is 0 Number of Time Units Specifies the number of time intervals to display in the chart For example if the data collection time interval is 10 seconds and the number of time interval units is set at 30 then up to 300 seconds or 5 minutes of data is displayed Low Threshold Enables and specifies a value line to be displayed on the chart Data below this line appears in the patterns and colors you set for low threshold High Threshold Enables and specifies a value line to be displayed on the chart Data above this line appears in the patterns and colors you set for high threshold Set Patterns and Colors To set patterns and colors for the selected instrument 1 Pull down the Modify submenu 2 Choose the Pattern and Colors menu item Performance Manager displays the appropriate Patterns and Colors dialog for that instrument Chapter 10 Customize the DECwindows Motif Real time Display 349 How You Edit the Panel Instruments Bar Graph The Bar Graph Patterns and Colors dialog allows you set the pattern and color of the instrument graph as shown in the following screen aph Pattems and Colors Graph Background Normal Range Low Threshold High Threshold Peak Hold ox eo cance ne
326. ee blocks required to set the data cell ANY DISK FULL to 1 for a given Performance Manager two minute record The maximum queue length on any disk required to set the data cell ANY DISK OVER QL THRESHOLD to 1 for a given Performance Manager two minute record This affects condition checking on rules 10040 10050 10055 10060 and I0150 Chapter 5 Customize the Knowledge Base 139 Components of Rules Name TD SMP VUP RATIO TD M0010 PROBRECS PER IMAGE TD PROCESS TD CIO TD IO ERROR TD CI PORT IO TD UNIBUS CHANNEL IO 140 Performance Manager Administrator Guide Default Value 85 1500000 1000000 Description A percentage of the threshold TD SINGLE CPU VUP n value applied to additional CPUs in an SMP system used to compute the total VUPs This is used to compute the data cell CPU VUP RATING for processor type N See the processor specific thresholds in the next table Number of records for any one image required to fire the rule M0010 Number of problem records per user image combination required to include the record into the evidence for rules M0050 M0055 and M0060 Number of problem records per disk volume required to include the image into the evidence for rule L0050 LOO60 and L0070 Number of problem records per disk volume required to include the volume into the evidence for rule LO080 Number of bytes per second through any CI port required to set the data cell EXCESS T
327. elect the data as a factor of the number of nodes being selected See Appendix D for information on estimating virtual memory needs and selecting data After selecting the data you can specify as many GRAPH PIE CHART or REPORT commands as you wish You can also select more data The chapter Using Command Mode Commands describes the commands available in command mode Because Performance Manager provides so many different choices for predefined and custom graphs you may prefer to use an interactive dialogue to make your selections 110 Performance Manager Administrator Guide Generate Multiple Graphs You can request this interactive prompting from within command mode by entering the GRAPH TYPE PROMPT command The Performance Manager s response is as follows PSPA gt GRAPH TYPE PROMPT Please select either 1 a predefined graph or 2 a custom graph Choice 1 Enter the graph type keyword lt cr gt for list cpu_utilization A graph is produced When you press Return prompting continues Generate Multiple Graphs When you generate multiple graphs with one DCL command or in Command Mode you can produce a separate output file for each graph by specifying the OUTPUT qualifier The Performance Manager names each file according to the node and graph type with a default or user specified file type For example the following command creates a separate graph output file for each graph The files reside in the default dire
328. em performance based on a large and changing body of technical information called a knowledge base This information is in the form of performance rules Performance rules come from one or two sources The first source of rules is the product itself The factory rules are automatically loaded by the user interfaces that require them The second source is an auxiliary rules file which you create While the Performance Manager is designed to work without any modifications an experienced system manager can further benefit from the tool by customizing the knowledge base Good reasons to customize the knowledge base include the following rules m Eliminating the firing of rules for conditions that are customary and unchangeable for your system m Refining rules to filter testing of specific images or other conditions that are a required and unchangeable part of your workload Modifying the characteristics of certain older disk devices that you might own and still use for example RF31 m Adding rules to check for additional warning conditions that are specific for your own workloads and systems Chapter 5 Customize the Knowledge Base 135 Investigate Rule Firing Investigate Rule Firing The factory rules embedded in the Performance Manager follow the methodology in the HP s OpenVMS Performance Management guide Some rules go beyond this methodology Keep in mind that while the Performance Manager alerts you to potential performance
329. en you have completed your set of changes to the file click Yes in response to this request Otherwise as long as your DECwindows Motif session remains active even though you may be doing other tasks such as graphing you still have the file locked A reminder of the status of the parameter file appears in the main window Chapter 8 Use the DECwindows Motif Interface 307 How You Customize Workload Definitions To create or modify workload definitions in the parameters file pull down the parameter submenu and choose the Workload Definitions menu item The Performance Manager displays the Workload Definitions dialog box as shown in the following screen Workload name New won oo i Include these processes Minimum WB interactive Bl Network base priority W batch W Detached Maximum base priority Workload is defined by sets of items Q Workload is unique by Account mane D Matching Either images or users Transn tion Image termination c3 requirement units Users Account Names LI The menu bar contains Control and Help menus A list of defined workload definitions appears at the top of the dialog box To close the dialog box pull down the Control menu and click the Exit menu item The Workload Definitions dialog box lets you do the following m Create a workload definition m Delete a workload definition Modify a workload definition
330. ents e 137 format e 137 source file e 172 S Save reports e 260 Schedule file display contents e 236 locked e 259 SCS e 85 Secondary keys e 34 Selection Filters Dialog Box e 269 Sequenced message e 85 Service time highest disk volume e 191 Set auto displays of panel instruments e 342 parameters e 319 data collector e 298 ranges and thresholds e 347 Severity code explanation e 403 Soft faults e 79 388 page fault in performance evaluation report e 34 Soft page fault in performance evaluation report e 34 Space used e 88 Specific data selection of e 267 Split I Os e 81 STACK keyword e 245 Standard process metrics e 90 Start DECwindows Motif Real time display e 321 341 DECwindows Motif interface e 257 Real time character cell displays e 374 Strip chart parts specify e 355 set automatic scaling e 348 high threshold e 348 low threshold e 348 maximum data value e 348 minimum data value e 348 minimum number of time units e 348 patterns and colors e 351 Super e 77 Swaper CPU e 77 SWAPPER e 75 System e 77 communication services e 85 metric for graphs e 126 generate a graph in command mode e 127 generate a graph in prompt mode e 127 overview panel e 324 display of e 341 working set field in system configuration report e 72 Systemwide DECnet Traffic graph e 191 Disk I O Rates graph e 191 Distributed Locking Activity graph e 191 Pagefaulting graph e 191 Terminal I O Rates graph e 191 T TABULAR keyword e 191 Tab
331. epresenting the number of balance set slots on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP GLOBALPAGE_FAULT_RATEMET_F_GVALID Average number of global page faults per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP GLOBAL_PGS_TALLY Derived This contains the sum of the values representing the number of global pages in the working sets for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO 452 Performance Manager Administrator Guide Performance Manager Data Cells HARD FAULT RATE Derived Average number of hard page faults per second for the local node for the current interval record This is derived from the sum of MET F PREADIO and MET F PWRITIO Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP HARD FAULT SCALING Derived This contains a value representing scaling factor for the hard page fault rate for the local node The value is obtained from the threshold TD HARD FAULT SCALING n where n is the hardware model number of the local node By default if the local node is a VAX 11 780 the value would be 1 0 The value of this data cell can be modified using a threshold construct Data Type NUMERIC Domains COM CFG CPU DSK FIL L
332. er Editor ADVISE EDIT command to set the AUTO AUGMENT parameter to the name of your rule file This setting becomes the system wide default The Performance Manager uses your auxiliary knowledge base file and augments the factory rules whenever you generate an Analysis Report or archive data When configuring AUTO AUGMENT to use an auxiliary rules file in an OpenVMS Cluster place the file in PSDC DATABASE or place it in another directory that is accessible by all nodes of the cluster If a logical name is used within the file specification be sure that the logical name is defined in the SYSTEM logical name table on all nodes Alternatively the cluster logical name table could be used Chapter 5 Customize the Knowledge Base 173 Use an Auxiliary Knowledge Base for Reporting and Archiving You can override the AUTO AUGMENT parameter and specify another auxiliary rules file with the following command ADVISE PERFORMANCE REPORT ANALYSIS RULES MY OTHER RULES KB You can override the AUTO AUGMENT parameter entirely and specify no auxiliary rules with the following command ADVISE PERFORMANCE REPORT ANALYSIS NORULES To clear the automatic augmenting of your site specific rules specify SET NOAUTO AUGMENT with the Performance Manager Parameter Editor 174 Performance Manager Administrator Guide Chapter 6 Performance Manager Commands This is a reference chapter for the Performance Manager command syntax It describes all comman
333. er Statistics Data Record Server Statistics Data Record The following example shows the server records that follow the Disk statistics shown in the Disk Statistics Data Record section in the exported file TYPE EXP SERV ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA HSCO HS70 1 0 2 0 0 0 0 0 0 Example HSCO HS70 1 0 2 0 0 0 0 0 0 Data item srv name srv type srv alloc cls srv work avail srv paging srv swping srv que length Description name type allocation class of time work was avail for server of server s work marked as pagio of server s work marked as swapio average length of server work queue Position 92 112 119 128 137 146 155 SERV Length 16 4 7 Appendix E Output Format for ASCII CSV Data 555 Process Metric Statistics Data Record Process Metric Statistics Data Record The following example commands show how to export and display process statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP PROC _ CLASS NODEFAULT PROCESS _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP PROC ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA PROC 26E00824 MACNEIL 00750 000021 4 LEF 0 0 0 0 0 0 0 0 0 0
334. ere soft resolved from memory without requiring a read from disk Note The bottom value in this column is the total page fault rate faults per second both hard and soft for the node 388 Performance Manager Administrator Guide Display RESOURCE Information Disk Display The Disk display contains disk statistics for analyzing an I O resource limitation in a cluster wide manner An example of a resource Disk display is shown in the following illustration JROUNDY COINA HABL IS R x A 5 M a IPPLE The top half of the disk display contains a bar graph for each disk currently set for display The scale shown at right and left shows an absolute value ranging from O to 100 The scale at left is headed with the word Rate while the scale at right is head Msec Milliseconds The rate scale is used with the leftmost column in the bar graph for a particular disk while the Msec scale is used with the rightmost column The name of the disk is given at the top of an individual disk graph Preceding the disk name is a number which corresponds to the number assigned the disk when listing disk groups with the SHOW GROUP command This number allows a partial disk name as given at the top of each disk graph to be associated with the full disk name as given by SHOW GROUP subcommand The number of I O packets in the disk queue for each disk is shown as a number at the bottom right of each disk graph Each disk graph has two columns as follows
335. erefore ignored Example Configuration Disk By Cluster Format NI 102 Performance Manager Administrator Guide MR 6031 AL Tabular Report Sections Disk By Node Format The following example shows disk statistics in By Node format Tabular Report CLUSTER PA Vx x Tuesday 26 JAN 2006 09 00 to 09 20 kakatakakatakalatakalatalalatakatatakakatakikok Final Statistics katakakatalakatakatatakakatalakatakatatakatok EERE EEE ALE LER EEE Data Analyzed from 26 JAN 2006 09 00 00 00 to 26 JAN 2006 09 20 00 00 EES RTE AER RE oe Avail Paging Swping Rate Time Queue Space 5 s ms Length Used DSA111 User Command ADVISE PERF REPORT TABULAR BYNODE BEG 26 JAN 2006 09 00 00 END 26 JAN 2006 09 20 00 00 0UT BYCLU TAB SECTION DISK In the By Node format for disk statistics the leftmost column of the tabular report contains the name of each disk followed by the name of each node from which the disk was accessed Following each node name is the processor type and if the node hosted the disk the word host Each disk with node names begins a line of disk statistics similar to the By Cluster format These statistics show the contribution of each individual node to the cluster wide activity of a specific disk The way the Performance software determines a node s contribution to the cluster wide activity of a disk depends upon whether the node has direct or remote access to the disk For nodes with remote access the
336. erformance Manager Administrator Guide Implement Changes TD HDW 10 TD RSR PPR 2 TD RSR INT 10 TD HIGH INTERRUPT STACK 20 TD HIGH KERNEL MODE 30 TD HIGH EXEC MODE 35 TD HIGH HARD FAULT 10 TD HIGH SOFT FAULT 100 TD HIGH DECNET RATE 100 TD IDLE MEM RATIO 0 05 TD HIGH SWAPPING 1 00 TD LOW SWAP IDLETIME 20 TD LOW BALSET MEM AVL 15 TD HIGH GLOBAL FAULT 40 TD HIGH SYSTEM FAULT 3 TD HIGH SPLIT IO 5 TD HIGH TURN RATE 6 TD FILE CACHE HITRATIO 70 TD FILE CACHE MISSEDIO RATE 5 TD COM PROCESSES 5 TD CPU NORMAL OVERHEAD 7 TD HIGH SYS BUFIO 70 TD HIGH FILE SYSTEM 30 TD HIGH SYS OPENS 5 TD HIGH TERM IO 60 TD HIGH IMGACT 0 5 TD POOL EXPANSION RATIO 0 40 TD POOL EXPANSION LIMIT RATIO 0 85 TD CIBCI PEAK KBTHRUPUT 1200 TD BCAA PEAK KBTHRUPUT 1400 TD BCAB PEAK KBTHRUPUT 2200 TD CIXCD PEAK KBTHRUPUT 9000 TD CI780 PEAK KBTHRUPUT 1900 TD ADAPTER SATURATION WARNIING RATIO 0 80 TD AVG MESSAGE SIZE KB 0 117 True 1 False 0 Endliteral While thresholds affect all rules literals affect the rules that are local to them that is included in the same file So to make this change effective you need to disable the original factory rules and copy over your own versions of this rule into your auxiliary knowledge base where it may be compiled with the new lower value of TD_HIGH_ERASE_IO Chapter 5 Customize the Knowledge Base 171 Build an Auxiliary Kno
337. erformance Manager Uses Virtual Memory see page 535 For Graphs see page 536 For Reports see page 537 For Integrity Servers and Alpha Systems see page 537 How Performance Manager Uses Virtual Memory Performance Manager uses process virtual address space to accumulate summarize and sort performance data for the presentation of graphs and reports The input data that is provided to PA for reporting or graphing determines the amount of memory required You may either count or estimate the number of items you have in the input data files to make an estimate for the amount of virtual memory needed Process data is generally the most likely category of performance data that occupies process virtual address space PA allows you to save process data at various levels of detail from process mode interactive batch network or detached to the most detailed level by process ID PID If you choose the most detailed level of reporting or graphing more virtual memory is required The following estimates assume a page size of 512 bytes On OpenVMS I64 and Alpha systems these would be considered pagelets Appendix D Estimate Virtual Memory Needs 535 For Graphs For Graphs To estimate the process virtual pages needed for ALL graphs m Use the following formula Pages required for graphing xp 128 nn 126 18 nu ni nw 28 nd 8 nf 2 nuid xp of x points nu of users of images
338. erformance Manager module This section contains the following topics PSPASDISPLAY PROCESS CPU UNNORMALIZED see page 405 PSPASDNS NAMES see page 406 PSPASEXAMPLES see page 406 PSPASGIVE DEVICE SERVICE see page 406 PSPASGRAPH CHARS see page 406 PSPASGRAPH FILE DEVICE see page 407 PSPASGRAPH FILE DIRECTORY see page 407 PSPASGRAPH LEGEND FONT POINT see page 407 PSPASGRAPH PATH see page 407 PSPASHLS see page 407 PSPASPIE FONT POINT see page 408 PSPASPS RGB 1 through PSPASPS RGB 6 see page 408 PSPASSKIP DISK FILTER see page 409 PSPASSKIP PIE PERCENT see page 409 PSPASSUPRESS TAPE STATS BY VOLUME see page 409 PSPASUNNORMALIZE CUSTOM CPU see page 409 PSPASDISPLAY PROCESS CPU UNNORMALIZED When this logical name is defined to anything the Real time Character Cell display utility ADVISE PERFORMANCE DISPLAY CHARACTER CELL displays the process CPU Utilization percentage relative to a single CPU rather than to the total CPU time available for all CPUs in the system This has an effect only when viewing processes on an SMP system containing more than one CPU By default all CPU percentages are displayed relative to the total CPU time across all CPUs in the system This logical name has an effect on the single node display and the process display when the current process key is either Top PID or Top Process Top Users Images or Accounts will always show the CPU
339. erisk indicates that a disk is a shadow set and the disk underwent a COPY operation Chapter 2 Analyze Performance 29 Brief Analysis Reports 5 Name of the node in the cluster that shares the heavily used disk 6 Average number of operations per second to the volume during the given interval by the contributing node 7 Average size of the queue during the interval exampled measured by the number of requests 8 Average size in pages of all I O requests during the interval 9 When hot file data exists the hottest file highest I O rate is listed In the previous Cluster Rule analysis report example the queue length on volume PROBLEM DISK exceeds the value of 1 0 on four occasions During those four occasions the total operations per second for each interval exceeded the device threshold Brief Analysis Reports The Brief Analysis Report is a synopsis of the Analysis Report It contains the following information m Rule identifiers m The percentage of time there were instances of rule occurrences during the reporting period This field is blank if the rule reflects an analysis of a summary of the over all analysis period m The number of records supporting the rule occurrence This field is blank if the rule reflects an analysis of a summary of the over all analysis period m A brief one line synopsis of the problem statement The Brief Analysis Report provides a synopsis for each node in the cluster system
340. erve a persistent performance problem you can use the Real time Motif displays to investigate dynamically see the chapters Use the DECwindows Motif Real time Display see page 321 and Customize the DECwindows Motif Real time Display see page 341 This new functionality supports progressive disclosure so that you can start monitoring with high level system displays and then progressively launch panels to focus on perceived problem areas as they are occurring Components of Rules A rule is a conditional statement in an if then format A rule can contain multiple conditions The Performance Manager evaluates rules while generating Analysis Reports If all conditions of the rule are true then there is a rule occurrence If the number of rule occurrences meet the rule occurrence threshold the rule is said to fire and the Performance Manager reports the associated rule conclusion A Performance Manager rules file contains entities known as constructs Performance Manager rule constructs represent rules Rule construct elements govern the syntax of the rule Rules File Constructs The format of a rules file is not rigid Multiple spaces tabs carriage return and form feeds are treated as a single space A rules file can include the following five constructs m RULE m DISABLE m LITERAL COMMENT m THRESHOLD Chapter 5 Customize the Knowledge Base 137 Components of Rules These rule constructs their format and use ar
341. es Enables graphs for listed categories m By Users Images of Volumes Enables graphing of process I O statistics by the user or image of volumes m By Selected Node Enables predefined top graphs and custom graphs for selected nodes The system metric graphs are always available By node This option greatly increases memory requirements m Additional Options Click the Additional Options button and the interface displays the Performance Manager Graph Options dialog box ll Auto Select Number of X Axis data points ta po Graph Averaging None OK Reset Cancel Help The Auto Select Number of X Axis data points button is used to have the Performance Manager choose a value that results in an even time interval to be represented by each data point When you disable this button you can specify the number of X Axis data points to plot across a graph As the value of X Axis data points increases spikes and valleys become more defined and the graph has a higher resolution A low number of X Axis data points produces a smoother graph because the graphing facility may average multiple data points within the time frame specified 276 Performance Manager Administrator Guide How You Select Data for Analysis Press on the Graph Averaging button and an option menu appears The selected data is averaged into the time period selected For more information on graph averaging see the chapter Generating Historical Gra
342. es The metric graphed is BUFIO TYPE TOP BUFIO WORKLOADS Top Buffered I O Workloads Plots the top five plus Other Users creators of buffered I O by workload names The metric graphed is BUFIO TYPE TOP BUSY DISKS Top Busy Disk Device Plots the five plus Other Disks disk devices that experienced the highest busy time percentages The metric graphed is BUSY TYPE TOP BUSY PROCESSOR Top Busy Physical Processor Plots the five plus Other processors that experienced the highest busy time percentages The metric graphed is P BUSY TYPE TOP BUSY VOLUMES Top Busy Disk Volume Plots the five plus Other Volumes disk volumes that experienced the highest busy time percentages The metric graphed is BUSY TYPE TOP CHANNEL IO Top HSC Channel I O Plots the five plus Other HSC channels that experienced the largest I O rate in I Os per second The metric graphed is CHANNEL IO TYPE TOP CHANNEL QUELEN Top Queue HSC Channel Plots the five plus Other HSC channels that experienced the largest queue length The metric graphed is CHANNEL QUELEN Note The channel names are provided in the format nodename n where n represents the channel number K SDI on the HSC node indicated by node name If the channel cannot be identified the character u is substituted for n See logical name PSDC hscname hscunitnumber in the Performance Agent Administrator Guide TYPE TOP CHANNEL THRUPUT Top HSC Channel Thrupu
343. ess they are averages weighted by the process s uptime m State Represents the scheduling state of the process for the most recent data For more than one process this column reflects the state of only one of these processes m Img Cnt Provides the number of image activations as a count for the process or processes m COM Prct Provides the percent of time that a process is observed in the COM scheduler state If more than one process is represented this figure could be greater than 100 however it is still a meaningful gauge as to which processes are queued up at the CPU m Sft Flt Sec Provides the average soft pagefault rate per second for the process or processes m Dsk I O Sec Provides the average number of disk I O operations per second for the process or processes m Tape I O Sec Provides the average number of tape I O operations per second for the process or processes m Process Mwait If the process or processes was observed by the data collector in an MWAIT scheduler state the resource name is provided RSN The low numbered resource states have precedence if more than one state was recorded The order of precedence starting with the highest is RWAST MAILBOX NPDYNMEM PGFILE PGDYNMEM BRKTHRU IACLOCK JQUOTA LOCKID SWPFILE MPLEMPTY MPWBUSY SCS CLU View Process Statistics with an Emphasis on IO Metrics The following report example illustrates the process statistics section with a focus on process IO related m
344. est as a single physically contiguous request and must split the logical request into multiple physical segments Usually Split I Os result from transfers occurring on fragmented disks Erase The number of disk erase I O operations per second for example when the DCL commands DELETE ERASE or PURGE ERASE are used File Opens This is the number of file open requests during the reporting interval AVE Open Files This section reports open files AVE Open File This is the average number of open files on any disk device during the reporting interval File Cache Attempt Rate This section reports the file system cache statistics System caches hold frequently accessed disk blocks of various types Blocks in file cache do not require disk I O therefore the use of caches expedites I O requests Dir FCB The number of attempts per second that were made to find directory file control blocks in the directory cache Dir Data The number of attempts per second that were made to find directory data in the directory cache Quota The number of attempts per second that were made to find entries in the quota cache 82 Performance Manager Administrator Guide Tabular Report Sections File Id The number of attempts per second that were made to find file identifiers in the file ID cache File Hdr The number of attempts per second that were made to find file headers in the file header cache Extent The number of attem
345. etrics primarily presented as rates Up to 3 top volume names associated with the IO activity are also displayed The example uses the default primary and secondary keys of USERNAME and IMAGENAME This format of the report is obtained by using the qualifier PROCESS STATISTICS FOCUS IO RELATED To view the report with a different orientation provide your choice of sort keys For example the following syntax presents the statistics by imagename with a breakdown by user PROCESS STATISTICS FOCUS IO RELATED PRIMARY KEY IMAGENAME SECONDARY KEY USERNAM E 46 Performance Manager Administrator Guide Performance Evaluation Report Performance YQUEM VAX 6000 440 Evaluation Tuesday 26 JAN 1997 09 00 to 10 00 Primary Secondary Key DirI O BufI 0 HrdFlt Img Dsk Ops KB Thru Top Disk 2nd Top Disk 3rd Top Dsk User Image Sec Sec Cnt Sec put sec Volume Name I0 sec Volume Name I0 sec Volume Name I0 sec ARROYO dcl TRAINING O SYSPACK COPY TRAINING O SYSPACK DECPRESENT TRAINING O SYSPACK DIRECTORY i i 3 SYSPACK MAIL i i TRAINING O SYSPACK BEYH dcl DECW QUOTE DOS CLIENT MATL NOTESSMAIN RTPAD BHAT dcl COPY DECPRESENT DIRECTORY MAIL QUOTE VO SET SUBMIT VMSHELP SYSPACK PAGER SYSPACK PAGER SWDEV SYSPACK SWDEV oooooo 2O2O0000 SYSPACK ADMIN ADMIN ADMIN SYSPACK SYSPACK ADMIN ADMIN SYSPACK SYSPACK SYSPACK SYSPACK eoooorco0oo cOococcococouvuo
346. etween the completion of a read I O request and the start of the next I O on the user s terminal for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 486 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS TERM RESPONSE TIME2PRO F RESPONSE TIME2 This contains a value representing the average number of milliseconds between the completion of a read I O request and the start of the next read I O on the user s terminal for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS TERM THINK TIMEPRO F THINK TIME This contains a value representing the average number of milliseconds between the start of a read I O request to the user s terminal and the completion of that I O for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS TERM THRUPUTPRO F TERM THRUPUT This contains a value representing the I O rate in bytes to terminal devices for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS UPTIMEPRO F UPTIME This contains a value representing the uptime in seconds of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS UPTIME TALLY Derived This contains the sum of the values representing th
347. event flag wait state Average number of processes in local event flag outswapped wait state Average process terminal response time for interactions requiring greater than 1 0 CPU seconds Number of local node ENQs or Lock Conversion CVTs per second Count of lock IDs in use Number of logical name translations per second Count of the LRPs in use Length of the LRP list Mailbox reads per second Mailbox writes per second Average process terminal response time for interactions requiring greater than or equal to 0 1 CPU seconds and less than 1 0 CPU seconds Chapter 6 Performance Manager Commands 213 ADVISE PERFORMANCE GRAPH Channel MEM TOTAL MFYCNT MFYFAULTS MODIFIED MP SYNCH MWAIT NETWORK COMQ NETWORK PROCESSES NP FREE BLOCKS NP FREE BYTES NP FREE LEQ 32 NP MAX BLOCK NP MIN BLOCK NP POOL MAX NV TERMIO OTHERBUFIO OUTGOING LOCKING OUTPROCACT 214 Performance Manager Administrator Guide Description Percent of physical memory in use excluding pages on the free and modified list Modified list page count Number of modified list pagefaults per second Percent of physical memory on the modified list CPU time charged while waiting for a resource protected by a spin lock to be freed Average number of processes in miscellaneous wait state Number of computable network processes Number of network processes Count of non paged blocks Number of free Kbytes in non pag
348. f metrics stored in the database that are selected with the SELECT command GRAPH The Performance Manager can produce a multitude of predefined graphs You can also define your own custom graphs if the predefined graphs do not meet your specific needs FORMAT Specifies the graph s output format Options include REGIS CHARACTERISTICS COLOR LINE PATTERN POSTSCRIPT CHARACTERISTICS COLOR LINE PATTERN TABULAR ANSI HEIGHT n LINE CSV The default value depends on terminal characteristics For more information see the section Advise Performance Graph see page 191 NODE NAME nodename Specifies the preparation of a graph for only one of the selected nodes The BY NODE graph processing option may be required during data selection if the metric is not a system metric By default the Performance Manager prepares graphs for all selected nodes Composite graphs Chapter 7 Use Command Mode Commands 245 GRAPH OUTPUT filespec Creates an output file that contains the graphs The default file extension for a ReGIS graph is REG the file type for ANSI and TABULAR formatted graphs is RPT and the file extension for PostScript is PS When you generate multiple graphs with a single command line you can create a unique output file for each graph To do this omit the file name with the OUTPUT qualifier The Performance Manager generates a separate file for each graph created and uses the graph type ke
349. f the specified strings The match string is case sensitive so if the process names have any lowercase letters spaces or tabs use double quotes when you enter the value e g FILTER PROCESSNAMES RTserver ACCOUNTNAMES Specify FILTER ACCOUNTNAMES to string chart all process records with the account name matching any of the specified strings WORKLOADNAMES Specify FILTERZWORKLOADNAMES workloadname to chart all process records associated with any of the specified workloads This filter is valid only if the CLASSIFY_BY qualifier is used to specify a classification scheme for your workload data UICS uic Specify FILTER UICS to chart all process records with the UIC matching any of the specified UICs An asterisk may be used to wildcard either the group or user field of the specified UICs PIDS pid Specify FILTER PIDS to chart all process records with the PID matching any of the specified PIDs Chapter 6 Performance Manager Commands 233 ADVISE PERFORMANCE REPORT Keyword Description NOLUMENAMES string Specify FILTER VOLUMENAMES to chart all disk records with the volume name matching any of the specified strings Do not specify any trailing colon DEVICENAMES string Specify FILTER DEVICENAMES to chart all disk records with the device name matching any of the specified strings Do not specify any trailing colon PERCENTAGE TOTAL MAXIMUM S
350. f the composite of selected nodes RESOURCE CNT Count of resources in use RT TERMIO Number of remote RT terminal I O operations per second SM RESPONSE Average process terminal response time for interactions requiring less than 0 1 CPU seconds SPLITIO Number of split I O transfers per second SRP CNT Count of SRPs in use Chapter 6 Performance Manager Commands 215 ADVISE PERFORMANCE GRAPH Channel Description SRP MAX Length of the SRP list SUPER Percent CPU time charged to supervisor mode SUSP Average number of processes in suspend wait state SUSPO Average number of processes in suspend outswapped wait state SYSFAULTS System page faults per second SYSTEMWS Percent of physical memory used by processes with the user name of SYSTEM TOTAL PROCESSES Total number of processes TRCNGLOS Transit congestion losses per second TT TERMIO Number of TT terminal I O operations per second TW TERMIO Number of DECterm I O operations per second TX TERMIO Number of TX terminal I O operations per second USERWS Percent of physical memory used by process working sets USER MODE Percent CPU time spent in user mode VMSALLOC Percent of physical memory allocated to OpenVMS including pool WINDOW TURN Number of file window turns per second WRTINPROG Transition page faults per second WT TERMIO Number of UIS terminal operations per second
351. faults total page faults second 91 7 0 0 pwrite fault pages write I Os second 100 7 S Appendix E Output Format for ASCII CSV Data 545 Page Statistics Data Record Example Data item Description Position Length 3 5 readio fault page read I Os second 109 7 s 28 9 pages read number pages 118 7 read second 0 0 pages writt number pages 127 7 en written second 33 5 dzro faults demand zero 136 7 faults second 56 8 gvalid fault global valid faults second 145 7 s 18 0 modify faul modified list faults second 154 7 ts 6 3 free faults free list faults second 163 7 0 0 sys faults system page 172 7 faults second 0 0 bad faults bad list faults second 181 7 0 1 trans faults transition state 190 7 faults second 3 0 hard faults of total faults which 199 7 were hard 97 0 soft faults of total faults which 208 7 were soft 0 0 write in pr write in progress 217 7 og faults second 546 Performance Manager Administrator Guide O Statistics Data Record O Statistics Data Record The following example commands show how to export and display I O statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP IO CLASS NODEFAULT IO _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP IO ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA IO 1 4 0 2 15 8 91 6 1066 0 1 2 13 8 97 0 45 4 3 2 3 2 0 0 Examp
352. fic hour point to the hour on the clock and click MB1 Holding down MB1 and dragging the pointer around the clock will turn off archiving for a series of contiguous hours Drag the slider on the scroll bar to display the archiving schedule for each day of the week or click the up or down stepping arrows Chapter 8 Use the DECwindows Motif Interface 317 How You Customize To duplicate a day s schedule 1 Press and hold MB3 inside the clock A pop up menu is displayed Release on the Cut menu item 2 Scroll to another day and press MB3 inside the clock Release on the Paste menu item The clock displays the copied schedule Workload Classification By default Performance Manager stores process data in the history file summarized by workload families To save modeling data in the history file click the model data unlimited button When model data is enabled no workload families can be selected and raw process data will be preserved Specific classification can then be done when the archived data is processed If you choose Classify by Families without specifying the workload families process data will be summarized into four records representing interactive batch network and detached processing All other process data will be lost such as process data based on image name account name and so on Workload Families Excluded The Workload Families Excluded list box contains a list of workload families To add a workload fami
353. file m Nofiles locked The schedule file is locked when you customize Data Collection and is unlocked when you complete your changes The parameters file is locked when you customize parameters When you complete your changes a message box appears asking you if you want to release the file See the section How You Customize see page 296 for more information about customizing either file How You Control the DECwindows Interface Pull down the Control menu and release on the menu item you want The Control menu lets you do the following actions m Save reports m Monitor work in progress m Read the Parameters file m Write the Parameters file m Load binary graph data m Save binary graph data m Quit the DECwindows session 260 Performance Manager Administrator Guide How You Control the DECwindows Interface Save the Reports To save the reports 1 Release on the Save As menu item to save reports Performance Manager displays the Save Reports dialog box The reports reflect the data selected for analysis as shown in the main window s Data Selected for Processing Section This option is disallowed if no data has been selected or if data selection was canceled For details of data selection see the section How You Select Data for Analysis see page 266 Chapter 8 Use the DECwindows Motif Interface 261 How You Control the DECwindows Interface gt 4 Save Reports Dialog Box Select Report Sections O Brief Analy
354. followed by a cluster wide synopsis Until you are familiar with the long version of the conclusions you should not rely solely on the Brief Analysis Report In many instances the one line synopsis is not sufficient to convey the meaning of the problem Interpret the Brief Analysis Report The following example is an example of a Brief Analysis Report A description of each item in the report headings follows the example 30 Performance Manager Administrator Guide Brief Analysis Reports The following statements are keyed to the columns in the report example 1 Rule identifier 2 Forthis reporting period the percentage of time that the conditions of a rule were satisfied 3 Number of records satisfying rule occurrences Rules in Summary domain do not provide data in columns 2 and 3 because their conditions are based on data averaged over the entire analysis period Brief Analysis CLUSTER Saturday 01 OCT 2006 00 00 to 23 00 NODE DEMAND ID oftime Recds One Line Description H Application program pagefaults very heavily M0500 A Heavy paging increase the working set extent for user s R0270 z Process es hung in AST See full report 10160 Window turns are too high alleviate file fragmentation R0070 More resources than hash table entries increase RESHASHTBL R0300 i Lots of contention for distributed locks Summary for DEMAND 6 rules fired of 337 records 8 satisfied conditions NODE SUPPLY ID oftime Rec
355. for the current tape record in TAPE domain e g MUA Data Type STRING Domains TAP TAPE_DEVNAMEMAG_A_DEVNAME This contains the OpenVMS device name string for the current tape record in TAPE domain e g 2 MUA1 Data Type STRING Domains TAP TAPE_ERROR_COUNTMAG_F_ERRCNT This contains the number of errors accumulated over the current interval for the current tape record in TAPE domain Data Type NUMERIC Domains TAP Appendix C Performance Manager Data Cells 519 Performance Manager Data Cells TAPE ERROR TALLY Derived This contains the sum of the ERROR counts for tape drives for all the tape subrecords which were selected by the most recent TAPE SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains TAP TAPE_IO_RATEMAG_F_OPCNT This contains the number of I O s per second for the current tape record in TAPE domain Data Type NUMERIC Domains TAP TAPE_IO_TALLY Derived This contains the sum of all I O rates to tape drives for all the tape subrecords which were selected by the most recent TAPE_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains TAP TAPE_SCAN Derived Provides the count of tape subrecords for which the specified rule condition is true The condition will be evaluated for each tape subrecord Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains TAP TAPE_SERVER_HWT
356. for the current interval Data Type NUMERIC Domains FIL FILE_SWAPPING_IO_TALLY Derived This contains the sum of the values indicating the number of swapping I Os per second transferred to and from all hot files indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME for the current hot file subrecords which were selected by the most recent FILE_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FILE_THROUGHPUTFIL_F_IOCNT This contains a value indicating the number of bytes per second transferred to and from the file indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME for the current interval Data Type NUMERIC Domains FIL Appendix C Performance Manager Data Cells 451 Performance Manager Data Cells FILE THROUGHPUT TALLY Derived This contains the sum of the values indicating the number of bytes per second transferred to and from all hot files indicated by FILE DEVICE FILE DIRECTORY and FILE NAME for the current hot file subrecords which were selected by the most recent FILE SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains FIL FREELIST_FAULT_RATEMET_F_FREFLTS Average number of page faults per second from the free page list for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP FREE_BALANCE_SET_SLOTS Derived This contains a value r
357. formance Manager Data Cells 411 Data Cell Navigation Table Data Cell Navigation Table Use the following commands to navigate the data cells Click r o nimjocoj oj oj Hi Sc x z l clia To move to this set of data cell descriptions ACTIVE PROCESSORS through AWSA IS SLOW BADPAGE FAULT RATE through BYTES IN PAGED POOL CHANNEL OVER THRESH PORT through CW VOLUME NAME DATAGRAMS DISCARDED through DYN MAXLEN ENQUEUE LOCKS NOT QUEUED RATE through EXEC FAMILY NAME through FREE BALANCE SET SLOTS GLOBALPAGE FAULT RATE through GLOBAL PGS TALLY HARD FAULT RATE through HSC TYPE HSC90 IDLE through IS A VAX KB MAPPED through KERNEL LARGEST BLK IN NONPAGED POOL through LRP MAXLEN MAILBOX READ RATE through MULTI IO NETWORK COUNT through NUM PROCS NOT USING WS LOANS OPEN FILES through OUT SWAP RATE PAGEFILE PAGE READ RATE through PRIORITY LOCKOUT PROCESSES IN CEF through PSWP WAIT QUOTA CACHE AR through QUOTA CACHE HR RDTS IN LIST through RDT WAIT RATE SCS ADAPTERNAME through SWAP WAIT SYSGEN ACP DINDXCACHE through SYSTEM FAULT RATE TAPE CONTROLLER through TROLLER IS ON USER through USER NAME VBS INTSTK through VOLUME NAME WINDOW TURN RATE through WS DECREMENTING TOO SEVERE XQP ACCESS LOCK RATE through XQP VOL SYNCH LOCK WAIT RATE 412 Performance Manager Administrator Guide Performance Manager Data Cells Performance Manager Data Cells ACTIVE PROCESSORS Derived This is the number of active CP
358. formance Manager also provides the capability to define your own rules User defined rules are identified using a text editor to create a file that contains rule definitions The chapter Customizing the Knowledge Base see page 135 explains the syntax of user defined rules In addition to writing new rules you can disable any factory rules After you define your site specific rules you must compile them before Performance Manager can use them The compiled version of your rules is called an auxiliary knowledge base After the rules have been compiled they can be used along with Performance Manager factory rules to create an Analysis Report You can have an auxiliary knowledge base used automatically or can specify it when requesting an Analysis Report Analysis and Reporting Facility The analysis and reporting facility generates the Analysis Report the Brief Analysis Report the Performance Evaluation Report the Tabular Report and histograms The Analysis Report identifies the effects of system parameter settings hardware configurations workload mixes and applications when they degrade the performance of individual nodes in the cluster or the entire cluster system Conclusions and recommendations are based on data collected The Brief Analysis Report is a synopsis of the Analysis Report It contains a one line description of each rule fired For more detail produce a full Analysis Report Chapter 1 Introduction 15 Performance M
359. ge faults for this type by the number of image activations A soft page fault is the total number of times that processes reference a virtual page that is not in its working set but is in memory Average number of hard page faults calculated by dividing the total number of hard page faults for this type by the number of image activations A hard page fault is the total number of times that processes reference a virtual page that is not in its working set and requires a read operation from disk Average number of direct I O operations per image Calculated by dividing the total number of direct I O operations for this type of process by the total number of image activations Chapter 3 Evaluate Performance in Detail 37 Performance Evaluation Report 22 Average number of buffered I O operations per image Calculated by dividing the total number of buffered I O operations by the total number of image activations Buffered I O operations use intermediate system buffers rather than process context buffers 23 Average CPU time used per image Calculated by dividing the total CPU time accrued by processes in seconds by the total number of image activations 24 Images per second Total number of image activations divided by the total elapsed wall clock time during which processes were active resulting in the average number of images completed per second Interpreting Process Statistics by Image Name and User Name The process statis
360. graphs e 126 info panel e 336 rate panels e 335 statistics e 88 DISKS graph type e 191 Disks panel e 335 Display interface types of e 178 Displaying reports brief analysis e 278 full analysis e 279 graphs e 286 edit of e 288 monochrome e 288 performance evaluation e 280 pie charts e 286 edit of e 288 process statistics e 282 selecting pie charts e 288 Domain element definition e 150 position within a rule construct e 148 Dzero faults e 79 Edit panel instruments e 345 ENQ e 84 Erase I Os e 81 Error messages displaying e 403 Evidence e 236 effects of batch or output file e 21 factory evidence e 155 format output e 154 rule element e 154 supporting evidence in analysis report e 21 Exec e 77 Expectations of the Performance Manager e 18 Extended process metrics e 90 Extent e 82 F Factory rules e 15 176 236 Fatal message e 403 FAULTS graph type e 191 Faults types panel e 331 File Index 579 cache effectivenesss e 83 cache hit and miss rates graph e 191 Hdr e 82 Id e 82 opens e 81 type e 177 G Generate lock request dequeue lock request e 54 Generate reports average CPU and memory e 62 batch procedure e 16 CI NI and adapter statistics e 55 cluster lock request headings e 54 conversion lock request e 54 CPU mode statistics e 51 CPU statistics e 44 daily e 16 disk activity e 56 headings e 56 hot file activity e 61 I O statistics e 46 lock traffic summary e 54 memory statistics e 48 pool statisti
361. greater than this number page faults might be excessive on your system Sysflts the rate at which pages are faulted into the system working set should be no more than 1 fault per second otherwise the system is faulting itself to do work on the users behalf If system faults are high it might be necessary to increase the value of system parameter SYSMWCNT which controls the system working set size If disks are spending an excessive percentage of their time doing paging indicated by the disk bars in the IO window then a memory limitation is causing harmful I O effects Swapping high Look at the value of Inswaps the value of Mem que and the amount of swapping and modified page writing done by the disks Swap Mod in the disk bars Inswaps gives the number of processes that were swapped back into memory during the last sample interval and Mem que is the number of processes waiting for memory If these are significant then swapping is a problem on your system and either indicates a memory limitation or memory management problem If disks are spending an excessive percentage of their time doing swapping and modified page writing indicated by the disk bars in the IO window then a memory limitation is causing harmful I O effects Investigate an I O Limitation The chief indicators of an I O limitation are as follows Direct I O high Look at the value of Dir I O as well as Data in the disk bar histogram Dir I O gives
362. gs establish the default colors for Performance Manager PostScript Graph Colors DEFINE PSPASPS RGB 1 22 1 55 Green DEFINE PSPASPS RGB 2 77 441 Magenta DEFINE PSPA PS RGB 3 88 77 11 Yellow DEFINE PSPASPS RGB 4 33 33 22 Brown DEFINE PSPASPS RGB 5 1 220 Red slightly Orange DEFINE PSPASPS RGB 6 001 Blue If you want to specify grey shades make the values for red green and blue the same For example DEFINE PSPASPS RGB 1 0 0 0 Black DEFINE PSPASPS RGB 2 2 2 2 DEFINE PSPASPS RGB 3 4 4 4 DEFINE PSPASPS RGB 4 6 6 6 DEFINE PSPASPS RGB 5 8 8 8 DEFINE PSPASPS RGB 6 95 95 95 near White 408 Performance Manager Administrator Guide PSPASSKIP DISK FILTER PSPASSKIP DISK FILTER For the disk reports you can enable this logical name to have entries for shadowset member units included By default shadowset member units are not included in the disk reports The logical name is enabled when defined to anything PSPASSKIP PIE PERCENT Define this logical to anything to suppress the printing of the percentages on the pie charts If the pie chart labels are so long that they extend beyond the sides of the paper the use of this logical name may make the labels fit PSPASSUPRESS TAPE STATS BY VOLUME The Tape Statistics section of the Performance Evaluation Report has a section for volumes and a section for tape devices If you define this logical name to anything the tape statisti
363. h e 191 Users graph e 191 Workloads graph e 191 Terminal Thruput Images graph e 191 Users graph e 191 Workloads graph e 191 Throughput Disk Device graph e 191 Disk Volume graph e 191 Files graph e 191 Images graph e 191 Users graph e 191 Workloads graph e 191 User Disk Operations e 191 Mode Processor e 191 Volume IO e 191 VA Space Images graph e 191 Users graph e 191 Workloads graph e 191 Workload Disk Operations e 191 Write Operations Disk Device graph e 191 Disk Volume graph e 191 Files graph e 191 WS Size Images graph e 191 Users graph e 191 Workloads graph e 191 TOP BDT W graph type e 191 TOP BLKS R graph type e 191 TOP BLKS S graph type e 191 TOP BUFIO IMAGES graph type e 191 TOP BUFIO USERS graph type e 191 TOP BUFIO WORKLOADS graph type e 191 TOP BUSY DISKS graph type e 191 TOP BUSY PROCESSOR graph type e 191 TOP BUSY VOLUMES graph type e 191 TOP CHANNEL IO graph type e 191 TOP CHANNEL QUELEN graph type e 191 TOP CHANNEL THRUPUT graph type e 191 TOP CLUSTER RULE OCC graph type e 191 TOP COMPAT PROCESSOR graph type e 191 TOP CPU IMAGES graph type e 191 TOP CPU RULE OCC graph type e 191 TOP CPU USERS graph type e 191 TOP CPU WORKLOADS graph type e 191 TOP CR W graph type e 191 TOP DGS D graph type e 191 TOP DGS R graph type e 191 TOP DGS S graph type e 191 TOP DIRIO IMAGES graph type e 191 TOP DIRIO USERS graph type e 191 TOP DIRIO WORKLOADS graph type e 191 TOP DISK
364. hat this process was seen at least once by the Performance Manager in the RWIMG mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWLCKPRO V RSN LOCKID This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWLCK mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWMBXPRO V RSN MAILBOX This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWMBX mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Appendix C Performance Manager Data Cells 491 Performance Manager Data Cells Domains PRO PROCESS WAS IN RWMPBPRO V RSN MPWBUSY This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWMPB mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWMPEPRO V RSN MPLEMPTY This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWMPE mwa
365. he local node s port to all other nodes for the current configuration record and interval Data Type NUMERIC Domains CFG PORT MESSAGES Derived This contains the value representing the number of messages sent and received per second from the local node s port to all other nodes for the current configuration record and interval Data Type NUMERIC Domains CFG PRIMARY IDLE Derived Average percentage of CPU time idle for the primary processor in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 476 Performance Manager Administrator Guide Performance Manager Data Cells PRIMARY INTERRUPT STACK Derived Average percentage of CPU time on the Interrupt Stack for the primary processor in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PRIORITY_LOCKOUT Derived This contains a Boolean value representing the presence of a priority lockout of a computable process by another with excessive CPU utilization on the local node for the current interval The process s priority which is causing the lockout must have a priority larger than DEFPRI Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PRIVATE_PGS_TALLY Derived This contains the sum of the values representing the number of physical private pages in the working sets for all of the current process subrecords which were select
366. he number of pages pagelets by which the limit of a working set is automatically decreased at each adjustment interval This value applies to the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN WSINCPAR F WSINC The value of the SYSGEN parameter WSINC which specifies the number of pages pagelets by which the limit of a working set is automatically increased at each adjustment interval for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 518 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN WSMAXPAR F WSMAX The value of the SYSGEN parameter WSMAX which is the maximum size of the process working set which determines the system wide maximum size of a process working set regardless of process quota for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSTEM_FAULT_RATEMET_F_SYSFAULTS Average number of system page faults per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP TAPE_CONTROLLERMAG_A_CTLR_NAME This contains hardware controller type string of the tape drive
367. he servers that provide access to the disks selected This qualifier is mutually exclusive with VOLUMES DNS_NAMES filename C Specifies the node name translation file for DECnet Phase V support when Phase V Node Synonyms are not defined This qualifier is used along with the MODE NETWORK qualifier The format of this ASCII file is one translation per line which consists of two names separated by a comma Provide the cluster name first and then the DECnet Phase V name or name segment or address For example LATOUR DEC TAY StanWilk For more information on this translation file see the appendix Performance Manager Logical Names see page 405 NODE NAMES nodename C W Specifies the node for which data is to be displayed By default all nodes associated with the collection definition are displayed MODE NETWORK DISKFILE C W Specifies which data gathering mode to use DISKFILE access does not start up a new data collector whereas the NETWORK keyword does DISKFILE allows use of the cluster s currently open collection files for the source of data DISKFILE cannot be used for displaying data from remote nodes Real time only for nodes within the current cluster However data from remote nodes can be selected for viewing in playback mode DISKFILE mode lets you display additional information that only the main CPD Performance Manager collects This includes process terminal response time hot file records an
368. he value of the SYSGEN parameter PQL DWSEXTENT which sets the default working set extent for a process created by the Create Process CREPRC system service or the DCL command RUN Process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PQL DWSQUOTAPAR F PQL DWSQUOTA The value of the SYSGEN parameter PQL DWSQUOTA sets the default working set quota for a process created by the Create Process CREPRC system service or the DCL command RUN Process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PQL MWSDEFAULTPAR F PQL MWSDEFAULT The value of the SYSGEN parameter PQL MWSDEFAULT which sets the minimum default working set size for a process created by the Create Process CREPRC system service or the DCL command RUN Process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 514 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN PQL MWSEXTENTPAR F PQL MWSEXTENT The value of the SYSGEN parameter PQL MWSEXTENT which set
369. hen the CPU is a limitation System CPU time high Compare the system CPU time sum of Int Ker Exe bars with the task CPU time sum of Sup Com Use bars The system CPU time is the sum of time spent on the interrupt stack and in kernel and executive modes The task CPU time is the sum of time spent in supervisor compatibility and user modes The sum of interrupt and kernel CPU time should not exceed 40 percent in most environments 396 Performance Manager Administrator Guide Evaluate Performance Using the Investigate Displays Isolate the Cause of a Memory Limitation If an examination of the system overview reveals a memory limitation you can investigate the cause of the limitation in more detail using the Memory display The following illustration is an example of the Memory display File Edi Commands Options Print GALLO a busses ents cans The Memory display draws attention to the following indicators Hard versus Soft faults Look at the value of Hard flt and Soft flt Hard flt gives the number of page faults per second that were resolved by reading from the disk Soft flt gives the number of faults per second resolved from memory A hard fault involves I O and is more expensive than a soft fault Hard faults in a properly managed system should be no more than about 10 percent of the total faults Hard flt Soft flt Inappropriate working set WS sizes Look at the Process bars at the right on the Memory dis
370. hin each panel Label Metric Units Next Panel Hard Fault Rate Hard page fault rate Faults per Fault Rates second Hard Fault Rate Hard page faults by user Faults per User faults for Top Users second list Total Memory Total memory utilization Percentage Memory Utilization Label Metric Units Next Panel Soft Faults Soft page fault rate Faults per second Hard Faults Hard page fault rate Faults per second Demand Zero Demand zero page fault Faults per rate second System Faults System page fault rate Faults per second In Swap In swap operations rate Swaps per second Out Swap Out swap operations rate Swaps per second Chapter 9 Use the DECwindows Motif Real time Display 331 Default Panel Descriptions User Faults Panel Process Faults Panel Label Metric Units Next Panel Hard Faults for Hard page fault rate for Faults per User this user second Hard Fault Rate Hard page faults by Faults per Process by User Process process for this user second Faults list Total Memory Memory for this user Pages x Pages Label Metric Units Next Panel Hard Faults for Hard page fault rate for Faults per Proc this process second User Name User name for this Alphabetic process Image Name Image name executed Alphabetic by this process Account User account for this Alphabetic process PID Process ID for this Numeric process Mode Process mode Alphabetic interactive batch network detached other CPU U
371. hosen Chapter 8 Use the DECwindows Motif Interface 295 How You Customize How You Customize To select and item for customization pull down the Customize menu and release on the item The Customize menu lets you perform the following actions m View and change Performance Agent settings m Specify or redefine the PSDC DATABASE definition m View and change Performance Manager parameters Customize the Data Collection To view Performance Agent settings wm Pull down the Customize menu and release on the Data Collection menu item The Performance Manager DECwindows Motif displays the Performance Agent Collection Definition dialog box The primary data collection process collects performance data according to parameters in the CPD collection definition within the schedule file If no other collection definitions exist the CPD parameters will be displayed in the Collection Definitions box as shown in the following screen Default working pu Def uult minimem 2500 set spate Defanit database reocsoatacase path Working Set Minimum Space Database Path co 20 FP50CTPATADASIE F30Ci9ATAhbAS E sae a l E CH meen A Bl wot file queue 0 330 Start date 22 SEP 1997 05 31 fles EZE day aer J Collection intervat 129 seconds Stop date manzano asan Schedule Legend B Collection Ow O Collection Off 296 Performance Manager Administrator Guide How You Customize
372. hown in the following screen Font Example ABC xyz 1234567890 Lox avoir cance rose neis The Font Selection dialog lets you specify the following font characteristics Family Size Weight Slant Chapter 10 Customize the DECwindows Motif Real time Display 353 How You Edit the Panel Instruments Note Font options not available on your server are disabled Metric Type Specifies a list of instances of one metric or a list of metrics For example to show the CPU Utilization for multiple processes on the system use the List of Instances option To show a CPU mode Interrupt Kernel Supervisor and so on for the system choose the List of Metrics option since each metric has only one value associated with it Number of List Entries Specifies the number of bars to be displayed Orientation Specifies the orientation of the bar graph s maximum value Data labels width in characters Specifies the space available for the data labels based on the characteristics of the font you have selected Meter width in characters Specifies the space available for the meter based on the characteristics of the font you have selected Set the toggle button to include a meter for the bars Specify the space for the meter in characters Units label width in characters Specifies the space available for the units label based on the characteristics of the font you have selected Set the toggle button to includ
373. hows a percentage value ranging from O to 10096 The name of the node is given at the top of an individual node graph while the average number of processes in the CPU queue for the node is shown as a number at bottom right There is a real queue if this number is greater than 1 This number is equivalent to the sum of processes in the computable queue COM Each node graph has two columns as follows 390 Performance Manager Administrator Guide System CPU S The INVESTIGATE Command This column is headed by the letter S and is a percentage value The value is not only given by the column height but is also shown as a number at the bottom of the column The value is the percentage of the total CPU time used by the System Interrupt stack Kernel mode Executive mode Task CPU T This column is headed by the letter T and is a percentage value The value is not only given by the column height but is also shown as a number at the bottom of the column The value is the percentage of the total CPU time used by user tasks Supervisor User and Compatibility modes Note that the bottom value in this column is the average number of processes in the CPU queue For a real queue to exist this value must be at least 2 The INVESTIGATE Command The purpose of the Investigate displays is to help you evaluate performance on one node All displays except one require use of a ReGIS compatible terminal You may use the keypad or IN
374. ht of the graph is 40 lines HISTORY history descriptor name Allows you to select history data from the Performance Manager database By default daily data files are used to supply data for graphing However by specifying the name of a history file descriptor you can select historical data instead You must define the history file descriptor in the parameters file and have archived data according to the descriptor s definition Use the DCL command ADVISE EDIT to invoke the Performance Manager Parameter Edit Utility From the utility you can ADD DELETE MODIFY and SHOW history file descriptors Use the ADVISE ARCHIVE command to create the archived files If history data with the periodicity attribute is selected the AVERAGE value is automatically set to that periodicity value This is true regardless of whether the AVERAGE qualifier is used 196 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH HISTORY is incompatible with the COLLECTION DEFINITION qualifier For information on how to produce a graph of history data including a typical time period see the chapter Generate Historical Graphs see page 107 Note If model data was not archived the CLASSIFY BY qualifier is restricted to those workload families specified in the history file descriptor NODE NAMES node name Identifies the nodes to graph The Performance Manager creates a separate graph for each node unless you specify the COM
375. ign 9 as the first character to indicate that the text is to be interpreted as a file name The default directory is assumed if not supplied as is a file type of DAT The format of the file must be a series of image names separated by white space or commas Supply only the filename field do not include the file type Image names can contain wildcard characters Image names can contain up to 39 characters If you wish to preserve lower case characters enclose image names in double quotes when you enter them 310 Performance Manager Administrator Guide How You Customize Users Press MB1 on the Users option menu to view the categories available Release MB1 on the entry indicating the type of user you want to create You cannot create a list until you make this selection To add a user field enter the appropriate string and press the Return key or click the Enter button The field will appear in the list box and be cleared from the text entry field To delete a user field click the entry in the user list box then click the Delete button The entry is removed from the list box To remove multiple entries click all entries to be deleted the fields and then click the Delete button A list of user entries can be provided through a file Use the at sign as the first character to indicate that the text is to be interpreted as a file name The default directory is assumed if not supplied as is a file type of DAT The format of the
376. ign 96 at the bottom of the column The value is the percentage of the CPU being utilized which is equal to the sum of the System and Task CPU percentages given in the tabular reports Chapter 11 Use the Character Cell Real time Display 387 Display RESOURCE Information Memory Display Hard Faults H Soft Faults S The Memory display contains memory statistics for analyzing a memory resource limitation is a cluster wide manner An example of a resource Memory display is shown in the following illustration The top half of the memory display contains a bar graph for each cluster node currently set for display The scale shown at right and left shows a percentage value ranging from O to 100 The name of the node is given at the top of an individual node graph while the total page fault rate faults per second for the node is shown as a number at bottom right Each node graph has two columns as follows This column is headed by the letter H and is a percentage value The value is not only given by the column height but is also shown as a number at the bottom of the column The value is the percentage of the total faults for the node that were hard required a read from disk This column is headed by the letter S and is a percentage value The value is not only given by the column height but is also shown as a number at the bottom of the column The value is the percentage of the total faults for the node that w
377. ile menu Chapter 9 Use the DECwindows Motif Real time Display 323 Use the Panel Commands Menu Use the Panel Commands Menu To specify commands Pull down the Commands menu and release on the menu item you want You can choose any of the following commands Connect Connects instruments to online data Disconnect Disconnects instruments from online data Set Interval Changes the update frequency of data in an instrument panel default of 10 seconds The interval is specified in minutes and seconds The instrument panel must be disconnected before this option is available Default Panel Descriptions The default panels within the Real time Display alert you to potential performance problems in any of the following four major resource categories CPU utilization CPU queuing Memory page faulting Disk I O activity When the Real time Display is started the System Overview default panel is displayed which provides a summary for these four performance indicators System Overview The System Overview Panel is the control panel for accessing information on any of these key system resources using the panel launch capability of the Real time Display The following table provides the panel information Label Metric Units Next Panel CPU Utilization CPU utilization Percentage 100 CPU maximum CPU Queue Computable queue Process count length 324 Performance Manager Administrator Guide Default Panel Descriptions Labe
378. ime spent in SUPERVISOR mode for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU SUPER TALLY Derived This contains the sum of the values representing the percent of time in SUPERVISOR mode for the current CPU subrecords which were selected by the most recent CPU SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU CPU_USERCPU_F_USER This contains the value representing the percent of time spent in USER mode for the physical CPU represented by the current CPU subrecord and interval Data Type NUMERIC Domains CPU CPU_USER_TALLY Derived This contains the sum of the values representing the percent of time in USER mode for the current CPU subrecords which were selected by the most recent CPU_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CPU CPU_VUP_RATING Derived This contains a value representing the VAX Unit of Processing VUP for a single physical processor for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 428 Performance Manager Administrator Guide Performance Manager Data Cells CW DISK CHANNEL IO Derived This contains a value representing the disk operations rate to the HSC s K SDI channel associated with the current disk for all nodes during the current interval Data Type NUMERIC Domains CLU
379. individual disks When generating a customized graph for a single metric with COMPOSITE the Performance Manager graphs the metric by node When graphing CPU percentages with the COMPOSITE qualifier each node s CPU time is scaled according to the VUP rating to produce a cluster average CPU utilization For more information see the chapter Generate Historical Graphs see page 107 DATES filespec Specifies that a file containing a series of date ranges is to be used in place of the BEGINNING and ENDING qualifiers Each line in the dates file should look like the following command dd mmm yyyy hh mm ss cc dd mmm yyyy hh mm ss cc The time can be either omitted entirely or truncated Any truncated parts of the time defaults to O The periods of time represented by each line in the file need not be contiguous but they must be in ascending order DATES is incompatible with the BEGINNING and ENDING qualifiers ENDING date Specifies the ending date and time of the graph Where date represents the date and time in standard DCL format If you do not specify BEGINNING ENDING defaults to the current time If you do specify BEGINNING the ENDING default are midnight 23 59 of the beginning date You can specify either an absolute time or a combination of absolute and delta times You can also use the keywords TODAY TOMORROW and YESTERDAY See HP s OpenVMS User s Manual or access the HELP topic SPECIFY DATE TIME for co
380. ing the presence of one or more processes which need larger working set maximums on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP PROCESSES_WAIT_IN_RWSWP Derived This contains a Boolean value representing the presence of processes waiting in a swapping file resource wait state on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 480 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS BASE PRIORITYPRO B PRIB This contains a value representing the process s base priority This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS BUFFERED IO RATEPRO F BUFIOS This contains a value representing the buffered I O rate of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS BUFFERED IO TALLY Derived This contains the sum of the values representing the buffered I O rate for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS COMMAND WAITPRO F COMMAND WAIT This contains a value representing the number of milliseconds elapsed from the start of the most recent terminal read request to the end of the current i
381. ing values m Interactive 30 m Batch 10 m Network 5 x Overhead 1 m Interrupts 5 m Other 0 If you specify PERCENTAGE MAXIMUM the pie chart contains an idle slice representing 49 percent of the total pie with the remaining 51 percent representing their respective slices If you specify PERCENTAGE TOTAL the pie chart represents the sum of these parts a total of 51 percent utilization with the largest slice of the pie approximately 3 5 being represented by Interactive If you do not specify PERCENTAGE on the PIE command line then PERCENTAGE MAXIMUM is assumed This qualifier has no effect on graphs custom pie charts or pie charts of metrics other than CPU Utilization Format Graphs and Pie Charts You can control the format of the Performance Manager s graphs and pie charts The following formats are available Format available Graphs Pie Charts PostScript Y y Tabular Y y CSV Y y ReGIS Y ANSI Y 116 Performance Manager Administrator Guide Format Graphs and Pie Charts The ANSI formatted tabular and PostScript output can be printed on any output device Color graphs print or display on monochrome devices in shades of gray See the discussion on the logical names to change the colors of the ReGIS and PostScript graphs in the Appendix Performance Manager Logical Names see page 405 Refresh a ReGIS Graph with New Characteristics After you generate a ReGIS graph to the SYSSOUTPUT device yo
382. ion in which the conclusion is reported See the following table An interval record contains data for a single node in a cluster system Domain Name Report Section Testing Frequency CLUSTER Cluster analysis Multiple times depending on the number of disks 150 Performance Manager Administrator Guide Components of Rules Domain Name Report Section Testing Frequency COMMUNICATION Local analysis Multiple times per interval record depending on the number of terminal controllers CONFIGURATION Local analysis Multiple times per interval record depending on the number of remote nodes within a cluster system CPU Local analysis Multiple times per interval record depending on the number of processors in a multiprocessor system DISK Local analysis Multiple times per interval record depending on the number of disk sub records FILE Local analysis Multiple times per interval record depending on the number of hot file sub records LOCAL Local analysis Once per interval record LOCALC Cluster analysis Once per interval record PROCESS Local analysis Multiple times per interval record depending on the number of process sub records SUMMARY Local analysis Once per node TAPE Local analysis Multiple times per interval record depending on the number of tape sub records Optionally you can display a trace each time a rule s condition is tested During report generation the Performance Manager displays the trace only on the terminal
383. ion process will hibernate until this date and time The formatting is as follows DD MMM YYYY HH MM End date The date and time on which a data collection process Is to terminate The formatting is as follows DD MMM YYYY HH MM 300 Performance Manager Administrator Guide How You Customize Changing the Collection Schedule A 24 hour clock is displayed A bar on the right side of the clock lets you scroll to each day of the week By default data collection is set ON for each hour of the day every day of the week To set the clock 1 Set or reset the square toggle button above the clock to turn data collection on or off for an entire day To turn off data collection for a specific hour point to the hour on the clock and click MB1 Holding MB1 down and dragging the pointer around the clock will set data collection to the value of the initial hour setting for a series of hours 2 Dragthe slider on the scroll bar to display the collection schedule for each day of the week or click the up or down stepping arrows To duplicate a day s schedule 1 Press and hold MB3 inside the clock A pop up menu is displayed Release on the Cut menu item 2 Scroll to another day and press MB3 inside the clock Release on the Paste menu item The clock displays the copied schedule Class Coverage Section The Class Coverage section lets you specify which major areas of performance data should be collected While the CPD can not be limited o
384. ion when calculating By Node and By Cluster disk statistics is based upon the node s relationship to the disk There are two types of relationships a node may have to a disk a disk may be either hosted or served by a node If a node is directly connected to a disk by a MASSBUS UNIBUS or HSC Hierarchical Storage Controller the disk is hosted by the node If a node is not directly connected to a disk and must go through an intermediary node that hosts the disk the disk is served For the purposes of this discussion the term direct access refers to the relationship where a node hosts a disk and the term remote access refers to the relationship where a node serves a disk The reason a node s relationship is important when calculating disk statistics is because the I O of a node with remote access to a disk is processed through a node with direct access to the disk Therefore rates for the node with remote access are included in disk statistics for the node with direct access and in the disk statistics for the remote node as well Chapter 3 Evaluate Performance in Detail 99 Tabular Report Sections The calculation of Total Cluster Disk I O rates is fairly simple the disk statistics from the nodes with direct access to the disk are added together Data from nodes with remote access to the disk is ignored as this data is already accounted for by nodes with direct access to the disk The following table is a summary of how the softw
385. ions SYSMAN UTILITIES EDITORS and so forth 40 Performance Manager Administrator Guide Performance Evaluation Report Performance YQUEM VAX 6000 440 Page 1 Evaluation PA Vx x Tuesday 26 JAN 2006 09 00 to 10 00 The table below lists observed workload characteristics of all the interactive images that were run during the given interval Note that Diskio Bufio and Cputim are percentage contributions of the respective images to the total system load Working set size and working set faults are the average for the respective images In the case of 0 image activations the Uptime image and Cputim image actually report the cumulative Uptime and Cputim for the image Node Name YQUEM WORKLOAD EDITORS of Page Faults Avg of of Uptime Cputim activ per Actvtn Ws Direct Buffered of image image Image ations Soft Hard size 1 0 I O Cputim EMACS LSEDIT TPU Node Name YQUEM WORKLOAD NETWORK of Page Faults Avg of of Uptime Cputim activ per Actvtn Direct Buffered of image image Image ations Soft Hard i 1 0 I O Cputim FILESERV LATSYM NETACP NETSERVER 44 REMACP RTPAD Node Name YQUEM WORKLOAD SYSMAN of Page Faults Avg of of Uptime Cputim activ per Actvtn Ws Buffered of image image Image ations Soft Hard size I O Cputim CONFIGURE CSP ERRFMT Chapter 3 Evaluate Performance in Detail 41 Performance Evaluation Report Performance YQUEM VAX 6000 440 Evaluation Tuesd
386. is functions This section contains the following topics Start the DECwindows Motif Interface see page 257 How You Control the DECwindows Interface see page 260 How You Select Data for Analysis see page 266 How You Display Analyzed Data see page 278 How You Customize see page 296 View the Main Window see page 320 Start the DECwindows Motif Interface To use the windowing interface the Performance Manager does not need to be installed or running on a workstation The windowing interface can be started and directed to your workstation by setting host to the node or cluster where the Performance Manager is installed and issuing the commands SET DISPLAY CREATE NODE mynode ADVISE INTERFACE MOTIF or ADVISE DECWINDOWS Note If the fonts required by the Performance Manager interface are not present a warning appears listing the expected font names All fonts used can be redirected to those available in your environment by modifying the Performance Manager resource DECPSSRESOURCES DAT For any font not available the DECwindows tool kit provides a best fit substitute which may alter the intended presentation For more information about the Resource File see the Installation Guide Chapter 8 Use the DECwindows Motif Interface 257 Start the DECwindows Motif Interface Use the Main Window When initiated DECwindows Motif displays its main window from which you select the activity you wa
387. is produced Each graph point value is determined by the sum STACKED of the items depicted up to 6 If the GREATER THAN keyword is specified without a value then 50 percent is assumed If the LESS THAN keyword is specified without a value then 90 percent is assumed If the SELECT qualifier is present without a keyword then GREATER THAN 50 is assumed For example ADVISE PERFORMANCE GRAPH BEGINNING 10 ENDING 11 NODE YQUEM _ TYPE CPU U THRESHOLD 25 CPU M THRESHOLD 35 TOP CPU I THRESHOLD 45 _ SELECT GREATER OUTPUT REGIS PSPA I CREAGRAPHOUT PSPA Graph created file BADDOG CORREY WORK PSPA YQUEM CPU UTILIZATION REG 1 This command requests that three graphs be produced The CPU Utilization graph is produced if 50 percent or more of the data points exceed 25 percent CPU utilization The CPU_MODES graph is produced if 50 percent or more of the data points exceed 35 percent CPU utilization The TOP_CPU_IMAGES graph is produced if 50 percent or more of the data points exceed 45 percent CPU utilization In this case only one graph is produced ADVISE PERFORMANCE GRAPH BEGINNING 10 ENDING 11 NODE YQUEM _ TYPE CPU_U THRESHOLD 25 CPU_M THRESHOLD 35 TOP_CPU_I THRESHOLD 15 _ SELECT GREATER OUTPUT REGIS PSPA I CREAGRAPHOUT PSPA Graph created file BADDOG CORREY WORK PSPA YQUEM CPU UTILIZATION REG 3 PSPA I CREAGRAPHOUT PSPA Graph created file BADDOG CORREY WORK PSPA YQUEM TOP CPU IMAGES REG 1
388. isk is not possible The By Node contribution for Node C which has remote access to Disk DU1 is the I O statistic for Node C For disk DB2 Node C is the host direct access node all I Os come to this disk through Node C Therefore to calculate Node C s By Node contribution subtract from Node C s I O information the I O information from Nodes A and B which have remote access to DB2 The By Node contribution of Nodes A and B is each node s I O statistics By Cluster and By Node Disk Statistics The following discussion explains the By Cluster and By Node Disk Statistics The disk statistics reported in By Cluster and By Node formats are as follows Work Avail9 o The percentage of time that work was available for the disk Reported only for By Node statistics Paging The percentage of the work available marked as Page I O Swping The percentage of the work available marked as Swap I O This includes both swapping and modified page writing Rate s The number of I O operations per second performed by the device 104 Performance Manager Administrator Guide Tabular Report Sections Resp Time ms The response time This is the mean total service time per I O request for the device in milliseconds This is the total time taken to service an I O request and is the sum of the queuing time and the busy time It is thus measured from the time the I O request is issued until the time the controller completes the reques
389. isplay 375 Display Multi node Statistics Command RESOURCE MEMORY DISPLAY RESOURCE RESUME RULES SET SCALING PROCESS n WORKING SET n RATE PER SECOND n USERNAME name Function Provides a multi node display of memory resources Provides the multi node resource display for the last selected display or the CPU display if none had been selected Resumes the data input stream that was stopped by the FREEZE command Provides a per node display of rules that have fired on that node Changes scale number within a tick mark for process working set size and I Os per second scales Provides a process display for the specified user name Display Multi node Statistics A bar graph style screen appears when you start Performance Manager character cell displays By default the percentage of CPU utilization for each node in the cluster is displayed If you are collecting data at two minute intervals factory rule IDs or user rule IDs or both may also appear after the time stamp For shorter intervals the user rules may appear See the RULES command and the RULES qualifier for more information 376 Performance Manager Administrator Guide Display Multi node Statistics You can display in the following metrics previous screen m Percentage of memory utilization m Disk I Os per second m Percentage of pagefile utilization m Any other system data cell collected and provided For a list of the sy
390. it state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWNPGPRO V RSN NPDYNMEM This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWNPG mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWPAGPRO V RSN PGDYNMEM This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWPAG mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 492 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS WAS IN RWPFFPRO V RSN PGFILE This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWPFF mwait state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN RWQUOPRO V RSN JQUOTA This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the RWQUO mwait state This cell pertains to the current process subrecord on the local node for
391. itation 0 000 0 397 Isolate the Cause of an I O Limitation 00000 ccc eee 399 Isolate the Cause of a CPU Limitation 000 000 nrnna anaana annann 400 Exit the Character Cell Displays 2 ls rh 401 Appendix A Performance Manager Messages and Recovery Procedures 403 Sample Performance Manager Message 22222 200 sss nn 403 Severity AA E 403 Appendix B Performance Manager Logical Names 405 PSPASDISPLAY PROCESS CPU UNNORMALIZED 222222222020200 a 405 PSPASDNS NAMES 2 222 22 nen en en en ene nent eens 406 PSPASEXAMPLES 22222222 eaaa aanroer narran araara n anaran 406 PSPASGIVE DEVICE SERVICE sssselseeless sees 406 PSPASGRAPH CHARS rrr rre EHE b pd der p de E E P deri i 406 PSPASGRAPH FILE DEVICE 222222 esses 407 PSPASGRAPH FILE DIRECTORY 222222200 sss 407 PSPASGRAPH LEGEND FONT POINT 22222200 sese 407 PSPAS GRAPH PATH 022222200 lesse 407 PSPAS HLS d REG GGG 499940 404900444045 BLDG 40 DINING NG RPG GG GU qu E RGAE SEDER SESE EE 407 PSPASPIE FONT POINT 22222 ese 408 PSPASPS RGB 1 through PSPASPS RGB 6 u 22 222 sss 408 PSPASSKIP DISK FILTER 2 2222 ess 409 PSPASSKIP PIE PERCENT 222222200 a 409 PSPASSUPRESS TAPE STATS BY VOLUME 222222220000 ess 409 PSPASUNNORMALIZE CUSTOM CPU 22222222200 sss 409 Appendix C Performance Manager Data Cells 411 Data Cell Navigation Table 2 sr 412 Performance Manager Data Cells 0
392. ived sec data gs sent data gram msg 120 7 sent sec data gs discd data gram msg 129 7 discarded k bytes mapd KB of buffer 138 7 space to receive send data from this node by local node k bytes requst KBytes of info to 147 7 receive from some remote node by local node Appendix E Output Format for ASCII CSV Data 549 System Communication Services Data Record Example 0 0 4 5 4 5 0 0 0 0 0 0 0 0 Data item k bytes sent msgs recvd msgs sent qd buf descrs qd buf credit reqst data sent data Description KBytes of info to send to some remote node from local node of msg received from this node by local node of msg send by local node to this node of times sec buffer descriptor entry not available of times sec local node had to wait for credits on remote node read ops sec initiated by local node for some remote node write ops sec initiated by local node for some remote node Position 156 165 174 183 192 201 210 Length 7 550 Performance Manager Administrator Guide Lock Statistics Data Record Lock Statistics Data Record The following example commands show how to export and display lock statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP LOCK CLASS NODEFAULT LOCK _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP LOCK ULTRA VERS Vx x REPORT
393. l Metric Units Next Panel Hard Fault Rate Hard page fault rate Faults per second Page Faults Disk IO second Total disk I O Operations per second Disk operation rate To navigate from one level of panels to the next double click a label shown on a panel Default Panel Hierarchy The following table shows the hierarchy of the default panels Resource Level 2 Panel Level 3 Panel Level 4 Panel CPU Utilization CPU CPU Modes User CPU Process CPU Image CPU CPU Queue Process Wait Image CPU Hard Fault Rate Page Faults Fault Rates User Faults Process Faults Memory Memory Allocation User Memory Page File Allocation Disk IO second Disk Volume Info Disk Info Each of the panels accessed from the System Overview panel are discussed in the following sections CPU Utilization Panel Descriptions The CPU Utilization graph on the System Overview panel lets you access the following panels m CPU m CPU Modes m User CPU Chapter 9 Use the DECwindows Motif Real time Display 325 Default Panel Descriptions CPU Panel CPU Modes Panel m Process CPU Image CPU This series of panels lets you investigate the use of the CPU on the monitored system You can investigate the top users of the CPU and view what processes and images each user is executing The following tables describe the information presented within each panel Label Metric Units Next Panel CPU Utilization CPU utilization Percentage CPU Modes
394. l node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP Appendix C Performance Manager Data Cells 527 Performance Manager Data Cells VMS83 OR LATER Derived This contains a one if the version is OpenVMS 8 3 or later zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP VOLUME_NAMEDEV_A_VOLNAME This contains a string indicating the volume label of the disk for which the current disk subrecord pertains in domain DISK Data Type STRING Domains DSK WINDOW TURN RATEMET F FCPTURN Average window turn rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP WORKING_SET_DEFAULTPRO_F_DFWSCNT This contains a value representing the UAF parameter WSDEF of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO WORKING SET DEFAULT TALLY Derived This contains the sum of the values representing the number of pages allocated as working set defaults for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL L
395. lates By Node disk statistics Disk Node Relationship Calculation of By Node Disk Statistics One or more nodes with Take I O data from node with direct access to direct access no nodes with disk remote access 100 Performance Manager Administrator Guide Tabular Report Sections Disk Node Relationship Calculation of By Node Disk Statistics One node with direct access Node s with remote access take I O data one or more nodes with from the node with remote access Node with remote access direct access subtract all remote access nodes I O statistics from the direct access node s I O statistics Information is unavailable if all data is not present that is when interval times for all nodes do not align More than one node with Unable to report By Node data direct access more than one node with remote access The following Example Configuration diagram below is an example system configuration There are two disks DU1 and DB2 and three nodes A B and C Disk DU1 is hosted by nodes A and B that is nodes A and B have direct access to DU1 Nodes A and B are served to disk DB2 that is nodes A and B have remote access to disk DB2 Node C has remote access to disk DU1 and direct access to DB2 Detailed examples of how the Performance software would calculate By Cluster and By Node disk statistics using the example configuration are given in the following By Cluster and By Node report format explanations Disk By Cluste
396. launch overrides the default You can also change the metric instance name by entering a new name in the Metric Instance text entry box The metric name is interpreted as all uppercase unless you contain the name within double quotes Include the metric instance name in the panel s title display Specify the name as the default Chapter 10 Customize the DECwindows Motif Real time Display 369 How You Set the Panel Options m Select prompting at connect time if a metric instance name is not defined m Define the prompt string For example if the instruments in a panel are set up to look at metrics for a specific user a user name is required to specify the metric fully In this case the prompt might be Enter User Name If prompting is selected then a prompt dialog appears at connect time requesting a user name Panels requiring specific metric instance names such as user name or process name must be provided with a prompt for the appropriate name This lets the panel to be invoked from the panel manager and to display without errors An example is the User CPU panel labeled User SAPIRO described in the Launching Panels section Note Any changes to the node selection or the metric instance name will have no affect on the instruments until the panel is connected or reconnected Remove Panel Menu To remove the panel menu 1 Pull down the Options menu and click the Remove Panel menu item The menu is removed from the panel
397. le Data item Description Position Length 1 4 w turns window turn 91 7 operations second 0 2 splits split I O 100 7 operations second 15 8 w hits window hits second 109 7 91 6 w hitr window 118 7 hits turns window hits 1066 0 openf number of open files 127 7 1 2 opens number of file open 136 7 ops sec 13 8 dirio direct I O 145 7 operations second 97 0 bufio buffered I O 154 7 operations second 45 4 lognam logical name 163 7 translations second 3 2 mbxread mailbox read 172 7 operations second 3 2 mbxwrites mailbox write 181 7 operations second 0 0 erase ios erase I O 190 7 operations second Appendix E Output Format for ASCII CSV Data 547 XQP Statistics Data Record XQP Statistics Data Record The following example commands show how to export and display XQP statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP XQP CLASS NODEFAULT XQP _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP XQP ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA XQP 99 7 3 0 100 0 0 1 100 0 0 2 100 0 0 3 80 8 5 5 88 7 4 5 0 0 0 0 Example Data item Description Position Length 99 7 dir hit directory FCB entries 91 7 found in cache 3 0 dir rate directory FCB cache 100 7 lookups second 100 0 quota hit percent quota entries 109 7 found in cache 0 1 quota_rate quota cache 118 7 lo
398. le Performance Manager Message The following illustration illustrates the parts of a sample Performance Manager message Severity Code Message Facility Identification PsPA Code Code Message Delimiter ed rt polum rrr erm 960 PSPA W DUPNOD duplicate node names entered Severity Codes The following table defines the severity codes that are assigned to messages Severity Code Explanation I Informational the Performance Manager software sometimes provides additional information about an action W Warning the command may have performed some but not all of a requested action verify the command or output Appendix A Performance Manager Messages and Recovery Procedures 403 Severity Codes Severity Code Explanation E Error The output or program result is incorrect but the Performance Manager software attempts to continue the execution F Fatal the Performance Manager software terminates execution of the request To display error messages m Type the following command HELP ADVISE PERFORMANCE ERROR Each description includes a recovery procedure Messages are listed alphabetically by the identification code that precedes the text of each message 404 Performance Manager Administrator Guide Appendix B Performance Manager Logical Names Performance Manager logical names begin with the prefix PSPAS This appendix lists those names and describes how they are used to control various aspects of the P
399. le the process statistics display from the Performance Evaluation Report use the following qualifier INCLUDE NOPROCESS STATISTICS Because process classification by PID or PROCESSNAME results in virtual memory requirements these reports keys are disabled by default and require you to specifically enable them For more information on virtual memory requirements see the appendix Estimate Virtual Memory Needs see page 535 34 Performance Manager Administrator Guide Performance Evaluation Report The following example shows a Performance Evaluation Report Process Statistics by Image for Interactive Batch Detached and Network Jobs Performance YQUEM VAX 6000 440 Page 1 Evaluation PA Vx x Tuesday 26 JAN 1997 09 00 to 10 00 The table below lists observed workload characteristics of all the interactive images that were run during the given interval Note that Diskio Bufio and Cputim are percentage contributions of the respective images to the total system load Working set size and working set faults are the average for the respective images In the case of 0 image activations the Uptime image and Cputim image actually report the cumulative Uptime and Cputim for the image Node Name YQUEM MODE INTERACTIVE of Page Faults Avg of of Uptime Cputim activ per Actvtn Ws Direct Buffered of image image ations Soft Hard size I O I O Cputim sec sec 142470 2 3 6 48 16 DEBUGSHR DECPRESENT 0 0
400. ll Analysis SUPPLY VAX 11 780 PA Vx x Saturday 01 0CT 2006 00 00 to 23 59 CONCLUSION 2 C0010 There is an apparent bottleneck at the CPU due to the large number of COM COMO processes There exists higher priority process es which are causing lower priority COM and or COMO processes to wait for the CPU which may be the cause of the problem This is considered a LOCKOUT condition Examine and or review process priorities For an equitable distribution of CPU time over the COM processes be sure they all have the same BASE PRIORITY Total number of samples giving this conclusion 4 CONDITIONS 1 COMPUTABLE PROCESSES GE 5 00 COM SCALING 2 PRIORITY LOCKOUT EQ 1 00 3 TOP CPU PROC CPU GT 7 00 4 OCCURRENCES GE 4 EVIDENCE Proc Process receiving most CPU COM Process in COM Time of or COMO USERNAME occurrence 1 OCT 00 04 00 1 OCT 00 06 00 1 OCT 00 08 00 1 OCT 00 10 00 The following statements are keyed to the columns in the previous CPU Rule analysis report 1 Average number of processes in computable or computable outswapped state during the interval 2 User name string of the process that consumed the most CPU time during the interval 3 The first 12 characters of the image name 4 Percentage of the total available CPU time consumed by the user s process during the interval 5 Base priority of the process that used the most CPU time during the interval 6 User name string for the process that w
401. llection were busy CPU IO Busy The percentage of time overlapped CPU and I O time that both the CPU and at least one disk device selected for data collection were busy The following I O statistics are reported in By Cluster and By Node formats Chapter 3 Evaluate Performance in Detail 97 Tabular Report Sections Direct I Os sec The number of direct I O operations performed per second during the example interval exclusive of page and swap I O Buffrd I Os sec The number of buffered I O operations performed per second during the example interval The following lock statistics are reported in By Cluster and By Node formats H Orig Lck Act sec This metric is the Host Originated Locking Activity per second This is the amount of locking activity generated by the host node It is equal to the sum of local and outgoing enqueue operations plus local and outgoing converted enqueue operations plus local and outgoing dequeue operations Out Bound 9 7 This is the percentage of the host originated locking activity above which had to be serviced by other nodes in the cluster It is equal to the sum of the outgoing enqueue outgoing converted enqueue and outgoing dequeue operations divided by the total host originated lock activity Enq Wait 2 0 This is the percentage of enqueue and converted enqueue operations that were forced to wait It is equal to the enqueue wait rate divided by the sum of loc
402. lobytes per device second Disk Disk read I O operations Operations per Reads second for this device second Disk Disk write I O operations Operations per Writes second for this device second Page Swap I O Rate Disk page and swap I O operations for this device Operations per second 336 Performance Manager Administrator Guide Review Data in Playback Mode Review Data in Playback Mode Playback mode lets you display data recorded earlier Select the Playback item from the File menu in the Panel Manager window Time A 16 FEB 1996 16 22 00 v Node Click the node for which you want to display data Time Time displays the time for which data is being displayed To select a time from which to display data records click Stop and then on the time box Edit the date and time to select the beginning time you want Click Play to start the display Alternately you can click the arrows to the right of the box to increment the time ahead or back Play Click Play to start a continuous display of the data from the time displayed Step Click Step to display data from the next interval Chapter 9 Use the DECwindows Motif Real time Display 337 Set the Thresholds and Ranges Set the Thresholds and Ranges You can set threshold values for strip charts and bar graphs to alert you to potential performance problems When a threshold value is exceeded the color of the indicator changes For example the
403. local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP ERASE_QIO_RATEMET_F_ERASEIO Average erase QIO rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP EXCESS_THRUPUT_ON_ANY_CHANNEL Derived This contains a Boolean value of zero or one one true represents the presence of excessive throughput on an I O channel on the local node for the current interval This is determined if the channel I O exceeds the amount indicated by the appropriate threshold TD MASSBUS CHANNEL IO TD UNIBUS CHANNEL IO TD KDA CHANNEL IO TD KDB CHANNEL IO or TD CI PORT IO Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP EXECMET_F_EXEC Average percentage of CPU time spent in Executive mode for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 446 Performance Manager Administrator Guide Performance Manager Data Cells FAMILY NAMEPRO A FAMILY This contains a string indicating the family name for which the current process subrecord pertains on the local node This filled in when the data is supplied from a history file otherwise it is blank Data Type STRING Domains PRO FASTER TERMINAL IO Derived This contains a value representing the sum of terminal operations rate to all but TTx terminals on the local node for the current interval Data Type NUMERIC
404. local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN COMOPRO V SSS COMO This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the compute outswapped queue This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 488 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS WAS IN CURPRO V SSS CUR This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in as the current process scheduled This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN FPGPRO V SSS FPG This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the free page wait scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS WAS IN HIBPRO V SSS HIB This contains a Boolean value where 1 represents the fact that this process was seen at least once by the Performance Manager in the hibernate scheduler state This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains P
405. ls m Disk m Volume Info m Disk Info This series of panels lets you investigate the I O activity on the monitored system You can investigate the top volumes and devices The following tables describe the information presented within each panel Label Metric Units Next Panel Disk IO second Disk I O operation rate Operations per second Top Volume Disk I O operation rate Operations per Volume Info list for top volumes second Top Device list Disk I O operation rate Operations per Disk Info for top devices second Label Metric Units Next Panel I O Rate for Disk I O operation rate Operations per Volume with for this volume second history I O Rate Disk I O operation rate Operations per current for this volume second KB second Disk throughput for this Kilobytes per s volume second Chapter 9 Use the DECwindows Motif Real time Display 335 Default Panel Descriptions Disks Info Panel Label Metric Units Next Panel Disk Disk read I O operations Operations per Reads second for this volume second Disk Disk write I O operations Operations per Writes second for this volume second Page Swap I O Disk page and swap I O Operations per Rate operations for this volume second Label Metric Units Next Panel I O Rate for Disk I O operation rate Operations per Device with second history I O Rate Disk I O operation rate Operations per current for this device second KB second Disk thruput for this Ki
406. ly name and the Workload Included list box display the definition of the selected family 2 Enter any of the following parameters Family Name When you modify a family name the Performance Manager assumes that you want to create a workload family based on the displayed definitions 314 Performance Manager Administrator Guide How You Customize Workload Specifications The Workloads Excluded list box contains a list of workload definitions that are not part of this family To add a workload click the workload definition name and click the Include transfer button right arrow To add multiple workload definitions click all their names contained in the Excluded list box Then click the Include transfer button To indicate a position within this included workload list click an existing entry in the Included list box All new workload definitions will be placed ahead of this entry To deselect a position entry click it again All new entries will be placed at the end The Workloads Included list box contains the names of the workload definitions in the family To remove an entry from the Included list box click the name and click the Exclude transfer button Control Buttons m The Clear button removes all entries m The Modify button applies the changes and clears the entries m The Delete button removes the selected family name and its definition m The Reset button redisplays all entries for the current family definition
407. ly to a history file definition click a workload family name and click the transfer button To add multiple workload families click all their names contained in the Excluded list box Then click the transfer button Workload Families Included The Workload Families Included list box contains a list of workload family names These names specify the workload families that define the new history descriptor To remove an entry from the Included list box click the name and click the transfer button Control Buttons m The Clear button removes all entries and cancels the definitions m The Create button adds the family definitions to the defined list and clears the entries 318 Performance Manager Administrator Guide How You Customize Delete a History File Descriptor To delete a history file descriptor 1 Click a defined history file descriptor name The dialog box displays the current definition values 2 Click the Delete button to remove the history file descriptor Performance Manager removes the descriptor name from the history file descriptor list and clears the definition fields If the history file descriptor is not deleted a message box displays explaining why your request was not executed A failure occurs when history files are created from this definition a list of these history file names displays 3 Delete the files then remove the definition Modify a History File Descriptor To modify a history file descriptor
408. m Click a defined history file descriptor name The dialog box displays the current definition values You can modify only the archive schedule Control Buttons m The Clear button removes all entries and cancels the modifications m The Modify button applies the changes and clears the entries m The Delete button removes a selected family definition m The Reset button redisplays the current family definition Parameter Settings To view parameter settings 1 Pull down the Parameters submenu and choose the Parameter Settings menu item Performance Manager displays the Parameter Settings dialog box as shown in the following screen Auto Augment Q on m gt Version Limit No Version Limit Q Version Limit days Chapter 8 Use the DECwindows Motif Interface 319 View the Main Window Highlighted buttons indicate the current settings Your changes in the Parameter Settings box are applied when you click OK or Apply Click the Reset button to restore the last settings that have been applied OR Click the Cancel button to close the dialog box without applying any changes From the Parameter Settings dialog box you can set any of the following parameters Auto Augment Click the ON button to establish automatic augmentation of an auxiliary knowledge base for analysis The initial setting is OFF When auto augment is set on the dialog box lets you enter the file specification of the compiled auxiliary rules fil
409. m Tape statistics m Hotfile statistics m Summary of node s CPU and memory statistics m Histograms of CPU and memory utilization and terminal and disk I O Chapter 3 Evaluate Performance in Detail 33 Performance Evaluation Report To display the Performance Evaluation Report specify the ADVISE PERFORMANCE REPORT PERFORMANCE EVALUATION command The FILTER qualifier lets you select a subset of data for reports For more information see the chapter Performance Manager Commands see page 175 Interpret the Process Statistics The following example illustrates the default process statistics section of the Performance Evaluation Report The PROCESS STATISTICS qualifier allows you to tailor the process statistics section of the Performance Evaluation Report You can specify the focus of the report to obtain different sets of statistics that pertain to the focus area The grouping merging and sorting of the process data is controlled with the primary and secondary key settings The following list shows the primary and secondary keys MODE m USERNAME IMAGENAME m UIC_GROUP m PROCESS NAME m WORKLOAD NAME m ACCOUNT NAME m PID For more information on how to specify the PROCESS STATISTICS qualifier see the chapter Performance Manager Commands see page 175 To display only the process statistics section of the Performance Evaluation Report use the following qualifier INCLUDE PROCESS STATISTICS To disab
410. mains PRO PROCESS_DIRECT_IO_RATEPRO_F_DIRIOS This contains a value representing the direct I O rate of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO 482 Performance Manager Administrator Guide Performance Manager Data Cells PROCESS DIRECT IO TALLY Derived This contains the sum of the values representing the direct I O rate for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS DISABLED ADJUSTMENTPRO B AWSA This contains a Boolean value where a 1 means the process has working set adjustment disabled This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS DISK IO RATEPRO F OPS This contains a value representing the disk I O rate of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS DISK IO TALLY Derived This contains the sum of the values representing the disk I O rate for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS DISK THRUPUTPRO F THRUPUT This contains a value representing the disk throughput of the current
411. mance Manager Real time and a complete description of all parameter options see the chapter Customize the DECwindows Motif Real time Display see page 341 The Real time Display shows the following default panel on your display System Overview on GALLO File Commands Edit Options Help CPU Utiization Hard Fault Rate 14 SEP 1993 21 04 00 2 00 If the primary Performance Manager is not running on the monitored system or the Real time data collector is unable to start an informational message is displayed The Performance Agent Administrator Guide describes the steps that the system manager needs to take to enable data collection for a user Control the Real time Display In addition to the initial default instrument panels an icon for the Panel Manager is displayed when you start the Real time Display To display the Panel Manager double click the Real time Display icon The Panel Manager is the control point for the Real time Display m To stop the display including the Panel Manager and all other panel windows that are invoked choose the Exit menu item from the File menu in the Panel Manager menu bar m To display previously recorded data choose the Playback item from the File menu 322 Performance Manager Administrator Guide Navigate Within the Default Panels The Panel Manager lists the panels available to you When you install Performance Manager for the first time the list contains the names of all the def
412. mand Mode Commands 239 SELECT Command LOAD GRAPH PIE_CHART REPORT SAVE PSPAWN EXIT HELP Function Loads a binary graph data file Causes a graph or pie chart to be produced from the selected data Causes the preparation of one of the reports Saves a binary graph data file Creates a subprocess of the current process Causes the program to exit Assists the user by providing a detailed discussion of any parameter or qualifier Executes the commands in the file spec SELECT The SELECT command selects data for analysis The SELECT command causes data to be selected for subsequent viewing by GRAPH and REPORT commands The following table lists all the SELECT You can abort the SELECT operation and return to the PSPA gt prompt by Format SELECT option Description command options entering Ctrl C Options Option ANALYSIS 240 Performance Manager Administrator Guide Function Enables the viewing of the Analysis Report from the selected data By default the Performance Manager provides analysis processing SELECT Option Function PERFORMANCE EVALUATION Enables the viewing of the Performance NO suboption Evaluation Report the Tabular Report and Histograms from the selected data Suboptions include HOT_FILE PROCESS key levels and ALL The PROCESS keyword may be followed by a list of PROCESS Key Levels indicating the detail level by which process data can be reporte
413. mber of mailbox writes performed per second during the reporting interval This section reports file system metrics XQP collected from OpenVMS performance statistics Window Hits The number of times the executive I O subsystem successfully maps a virtual to logical segment without needing to invoke XQP services Window Turns The number of times the XQP updates the Window Control Block WCB Window turns occur when the executive I O subsystem fails to map a virtual to logical segment using the current contents of the Window Control Block WCB The XQP updates the WCB with virtual to logical mapping information by reading a new portion of the file s header from disk or cache and reissues the I O transfer A large number of window turns usually indicate that a file or volume is fragmented If the WCB is regarded as a cache of file mapping pointers each window turn indicates a cache miss Chapter 3 Evaluate Performance in Detail 81 Tabular Report Sections A very large file may cause excess window turns due to its size even if the file is contiguous This is because the maximum size of a window control block pointer is 65K blocks If you encounter this case you should provide a larger default window size when mounting the disk Split I Os The number of times the executive must map and queue a segment in a multi segment request to a driver A split I O occurs when the executive I O subsystem cannot map a single logical I O requ
414. mily definitions history file descriptors and auxiliary knowledge base information See the Performance Agent Administrator Guide for more discussion of these definitions Start Parameter Editing To edit parameters 1 Pull down the Customize menu in the Performance Manager Main Menu 2 Click the pointer on the Parameters menu item 3 Choose a menu item from that submenu Only one user at a time is allowed to edit data in the parameter file PSDC PARAMS DAT The file is locked by anyone using any editor If the file is locked by another user when you initiate any of the parameter menu s submenus a message box appears explaining that read only access to the file is allowed You are able to view the contents of the file but any requests to change data are denied Should the file become unlocked during the course of your DECwindows Motif session a second message box appears asking you if you would now like update capabilities Indicate your choice by clicking on either Yes or No 306 Performance Manager Administrator Guide How You Customize Once you have gained update access to the file it is unavailable to other users Whenever you close one of the parameter editor s dialog boxes you are asked if you would now like to release the file Specify PSDC Database Directory Equivalence name s USERSER6 DECPS DATARASE Logical Name Table O Process table Job table Group table System table ex CJ Go pev Wh
415. ministrator Guide Create a History File Descriptor Enter any of the following parameters History Descriptor Enter the name of the history file descriptor The descriptor name is limited to 20 characters Data Reduction Scheme By default the granularity value is monthly Click and hold the pointer on monthly and the granularity pop up menu displays a list of value options Release the mouse button when the cursor is on your choice The pop up menu disappears and your selection is displayed By default the periodicity value is None Click and hold the pointer on the current value and the periodicity pop up menu displays a list of value options Release the mouse button when the cursor is on your choice The pop up menu disappears and your selection is displayed By default the interval value the time period over which the Performance Manager averages daily data records into a single history data record is 60 minutes To display valid entries click the up or down arrow When your selection is displayed click the value Your choice will be highlighted Archive Schedule A 24 hour clock is displayed A bar on the right side of the clock lets you scroll to each day of the week By default archiving is set on for 24 hours a day seven days a week including holidays To set the clock 1 Set or reset the square toggle button above the clock to turn archiving on or off for an entire day To turn archiving off for a speci
416. ministrator Guide Histogram History database History file A histogram is a ASCII graphic chronological chart showing resource use The history database is reduced data from the daily data files which resides in the history files The history file contains data archived from the daily data files The number of history files created and maintained depends upon the number of history file descriptors and the associated granularity The filename has the form PSDC DATABASE PSDC nodename dd mmm yyyy dd mmm yyyy name History file descriptor Holidays file Hot files Interaction Interval Journal file The history file descriptor contains the description that Performance Manager uses to determine how to archive data to the file The holidays file contains a list of holiday dates The Performance Agent uses the holiday schedule on these dates The file name is PSDCSHOLIDAYS TXT Hot files are the most frequently accessed files on each disk The Performance Agent collects hot file data when the length of a disk queue exceeds the HOTFILE QUEUE setting An interaction with a device is a two step process First a unit of work to be done job user process and so forth enters the queue of the device Then the unit of work is serviced by the device in a manner dependent on the queueing discipline and departs the device The concept may be generalized to the system as a whole The set of device interactions required to process
417. ministrator Guide Performance Manager Data Cells WORKING SET QUOTAPRO F WSQUOTA This contains a value representing the current working set quota of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO WORKING SET QUOTA TALLY Derived This contains the sum of the values representing the number of pages allocated as working set quotas for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO WORKLOAD NAMEPRO A WORKLOAD This contains a string indicating the workload name for which the current process subrecord pertains on the local node This filled in when the data is supplied from a history file otherwise it is blank Data Type STRING Domains PRO WRITE IN PROGRESS FAULT RATEMET F WRTINPROG Average number of write in progress page faults per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP WS DECREMENTING NEEDED Derived This contains a Boolean value detecting a condition warranting working set decrementing on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 531 Performance Manager Data Cells WS DECREMENTING TOO SEVERE Derived This contains a Boolean
418. mode for one or more nodes in a cluster system RESOURCE CPU DISPLAY RESOURCE MEMORY DISPLAY RESOURCE DISK DISPLAY RESOURCE or press PF3 Chapter 11 Use the Character Cell Real time Display 385 Display RESOURCE Information RESOURCE Keypad The following illustration shows the functions of the RESOURCE keypad Invest igate KEYPAD DEFINITION Display for the RESOURCE DISPLAY Next and Prev Screen Allow you to view the various RESOURCE displays CPU MEMORY and DISK Zoom Out to Z Exit PSPA RT Multi Node Display Balance Cluster System Utilization Using the Resource Display Use the Resource displays to determine the workload on nodes and disks on your cluster system and balance the workload as necessary For example if your Resource display shows one node that has a high percentage of a resource in use you may wish to move work from that node to other nodes to balance resource utilization on your cluster system Each display shows cluster wide information consisting of memory disk or CPU metrics for selected nodes and disks in a cluster Each display consists of the following two parts m An upper part containing metrics appropriate to the display name for example memory related metrics This part is unique for each different type of display A lower part containing memory utilization direct I O rate and CPU utilization for selected nodes in the cluster This part is the same for each of the three
419. modeling the interval between the moment a request arrives at a device and request completion at a device Glossary 571 Rule Rule conclusion Rule condition Rule elements Rule evidence Rule expressions A rule is one or more rule conditions that are evaluated when Performance Manager Analysis reports are generated Rules are applied to daily or historic data If all conditions for a rule are true then there is a rule occurrence Rules are defined to expose areas of potential system problems See Rule firing A rule conclusion is a rule element The conclusions are Performance Manager recommendations based on the conclusion text element of a rules file rule construct A rule element A rule condition is made up of one or more rule expressions Describes the circumstances that must be true to cause a rule occurrence The seven rule elements that can exist in a rule construct are as follows Brief Conclusion element Conclusion element Domain element Evidence element Occurrence element Rule condition element Rule ID element A rule element Data satisfying a rule occurrence in a Performance Manager analysis report The evidence consists of data cell names and values Typically these data cells are some of those contained in the rule expressions Components of rule conditions which may include the following 572 Performance Manager Administrator Guide Rule firing Rule identifier Rule occurrence Rules compiler
420. mplete information on specifying time values ENDING is incompatible with the DATES qualifier Chapter 6 Performance Manager Commands 193 ADVISE PERFORMANCE GRAPH FILTER keyword The FILTER qualifier allows you select a subset of the daily or history data for graphing Process data and disk data can be filtered Hotfile data is also filtered When you specify filtering by process a hotfile record is selected if accessed by the specified process When you specify filtering by disk device a hotfile record is selected if located on the specified device For hotfile records matching both process and disk device specify filtering by both process and device Process data can be filtered by using any of the filter keywords USERNAMES IMAGENAMES PROCESSNAMES ACCOUNTNAMES UICS PIDS or WORKLOADNAMES If a process record s identification information matches any of the identification specifications that are specified then that record is selected Likewise disk data can be filtered by using either of the filter keywords VOLUMENAMES and DEVICENAMES If a device record s identification information matches any of the volume names or device names that are specified then that record is selected The following table lists the FILTER keyword options Keyword USERNAMES string Description Specify FILTER USERNAMES to graph all process records with the username matching any of the specified strings IMAGENAMES
421. n absence of the Performance collection process Equivalent to the sum of processes in the computable queue COM 76 Performance Manager Administrator Guide Swapper Counts CPU Statistics Tabular Report Sections This section reports the metrics of the swapper process InSWP The number of process inswaps performed during the reporting interval OutSWP The number of process outswaps performed during the reporting interval Header The number of process headers swapped in during the reporting interval Header OutSWP The number of process headers swapped out during the reporting interval A process body may be swapped out without outswapping the corresponding header Swapper CPU The percentage of CPU time used by the SWAPPER process This includes time for swapping modified page writing and process working set trimming activities Also some of the swapper activity may be reported as MP_SYNCH time In a multiprocessor system the statistics for the additional processors are included for the following metrics CPU ID A unique number distinguishing one processor from another Total Idle The percentage of time that the CPU was idle MP_SYNCH Wait The time a CPU spends waiting to acquire a spinlock in kernel mode This metric is always O on a single processor system Inter Stack Percentage of CPU time spent executing on the interrupt stack Kernel Percentage of CPU time spent executing in kernel mode for example
422. n a multiprocessor system both CPUs must be idle Page or Swp Wait Percentage of time that the CPU was idle and at least one disk device had either page I O or swap I O in progress In a multiprocessor system both CPUs must be idle Page and swap data is based on statistics Performance collects at 100 millisecond intervals from I Os waiting to be processed 78 Performance Manager Administrator Guide CPU and I O Overlap Paging Rates Tabular Report Sections This section reports the CPU and I O overlap statistics collected by the PSDCTIMER EXE timer driver CPU IO Idle The percentage of time that the CPU and all disk devices were idle CPU Only The percentage of time non overlapped CPU time that a CPU was busy and no disk device was busy I O Only The percentage of time that the CPU or all CPUs in a multiprocessor system were idle and at least one disk device Multi I O The percentage of time that two or more of the disk devices were busy CPU IO Busy The percentage of time overlapped CPU and I O time that both the CPU at least one CPU in a multiprocessor system and at least one disk device were busy Note CPU and I O overlap statistics are not available if the PSDCTIMER EXE driver was not loaded when data was collected This section reports paging subsystem or memory management metrics This data is collected from OpenVMS performance statistics Page Faults The total number of page faults per second bo
423. n in following example the first three elements rule ID domain and the rule conditions must be in this order The next four elements may be in any order disable M0010 DISABLE M0010 fire this rule after 10 rule occurrences LITERAL Too many 10 ENDLITERAL Rule UM010 Domain Local Soft fault Rate ge 100 Hard fault Rate ge 8 Pages on Freelist le 300 Computable Processes Ovr Defpri ge 2 5 Direct IO Rate ge 40 Occurrences Too many Evidence Soft Fault Rate Hard fault Rate Pages on freelist 148 Performance Manager Administrator Guide Rule ID Element Components of Rules Computable Processes Ovr Defpri Direct IO Rate User name Top HF User X Volume Name Highest IO rate disk x Time Conclusion There are significant demands on all of the system s resources Either lower the overall demand or expand the data processing resources Brief conclusion System resources fully taxed performance degradation likely EndRule The rule ID identifies the rule Each rule must begin with a unique rule identifier The format of a rule identifier is this RULE rule ID Where rule ID m Is one to five alphanumeric characters m Does not have a zero for the second character The zero is reserved for use by CA m Is not already defined by Performance Manager factory rules By convention the first character of a rule identifier describes the rule performance category The Performanc
424. n the associated button on this dialog The Pattern Editor dialog is displayed 5 Select new patterns and colors for each graph part click OK on all dialoges to effect the change 340 Performance Manager Administrator Guide Chapter 10 Customize the DECwindows Motif Real time Display This chapter provides information about using the Panel Manager to customize the DECwindows Motif Real time Display This section contains the following topics Access the Panel Manager see page 341 Specify Actions on Panels see page 342 Terminate the Session see page 345 How You Edit the Panel Instruments see page 345 How You Set the Panel Options see page 366 Access the Panel Manager Use the following steps to access the Panel Manager 1 Start the DECwindows Motif Real time display by entering the following command ADVISE PERFORMANCE DISPLAY WINDOWS MODE NETWORK For more information about the DISPLAY WINDOWS command syntax see the chapter Performance Manager Commands Performance Manager displays the Panel Manager icon and the System Overview panel as shown in the following screen Equation 1 System Overview Panel System Overview on GALLO File Commands Edit Options CPU Utilization Chapter 10 Customize the DECwindows Motif Real time Display 341 Specify Actions on Panels 2 Double click the Panel Manager icon or from the System Overview Panel pull down the File menu and click the Pa
425. n the dashed outline defining the field 5 When you are satisfied with all the selections click OK Chapter 10 Customize the DECwindows Motif Real time Display 363 How You Edit the Panel Instruments Launch Panels Once launch panels are set and you are connected double click the metric name or metric instance name you want to pass to the specified launch panel Performance Manager displays this panel using the information passed For strip charts digital meters and bar graphs with one or more different metrics displayed List of Metrics as the type you can now launch the selected panel by double clicking on the instrument title For bar graphs that display multiple instances of a metric for example a bar graph of the CPU utilization for the top 7 users you can double click an instance identified by user name and have that name passed to the launched panel The launched panel can then filter its displayed metrics based on the specific user See the section Assigning Metrics see page 359 for information on metric filtering To disable panel launching select the Remove Launch Panel menu item and click OK The following example illustrates progressive disclosure using panel launching f System Overview on YQUEM File Commands Edit Options Help CPU Utiizaton Hard Felt Rate 364 Performance Manager Administrator Guide How You Edit the Panel Instruments In the following example double clicking on
426. nager displays the Panel Title dialog as shown in the following screen Panel Title PSPA RealTime on 2 Enter the title and click the OK button The supplied string can optionally have appended to it the node name and the instance value 368 Performance Manager Administrator Guide How You Set the Panel Options Specify the Panel Node and Metric Instance Data To specify the node for which data is to be displayed 1 Pull down the Options menu 2 Choose the Node and Metric Instance Name menu item Performance Manager displays the Node and Metric Instance Name dialog as shown in the following screen Bl incisde Node in Diplay O Save as Default Node Metri instante Name lc d C inctede Name in Display LJ Save as Default Instance C frempt for Undefined Name apay Cj aes s The Node and Metric Instance Name dialog lets you perform the following actions Show the current node assigned to the panel This would have been specified by you as a default or at connect time or when the panel was launched If the panel was launched the node passed in the launch overrides the default Include the node in the panel s title display Specify the node as the default Show the current metric instance name This would have been specified by you as a default by a prompt dialog at connect time or when the panel was launched If the panel was launched the metric instance name passed in the
427. name of the remote node in the cluster system configuration for the current configuration subrecord Data Type STRING Domains CFG SCS NODE HWNAMECFG A HWNAME This contains the hardware type string of the remote node for the current configuration subrecord Data Type STRING Domains CFG SCS NODE IS HSCCFG V STATUS HSC This contains a Boolean of either the values 1 if the hardware type of the remote node for the current configuration subrecord is a HSC or a 0 if it is not an HSC Data Type NUMERIC Domains CFG 496 Performance Manager Administrator Guide Performance Manager Data Cells SCS NODE IS MEMBERCFG V STATUS MEMBER This contains a Boolean of either the values 1 if the remote node for the current configuration subrecord is a cluster member or a 0 if it is not a member Data Type NUMERIC Domains CFG SCS NODE IS VAXCFG V STATUS VAXNODE This contains a Boolean of either the values 1 if the remote node for the current configuration subrecord is a VAX or a O if it is not a VAX Data Type NUMERIC Domains CFG SCS NODE ON CICFG V STATUS CI This contains a Boolean of either the values 1 if the remote node for the current configuration subrecord is accessed over the CI or a 0 if it is not Data Type NUMERIC Domains CFG SCS NODE ON NICFG V STATUS NI This contains a Boolean of either the values 1 if the remote node for the current configuration subrecord is accessed over the NI or a 0 if it
428. nce Manager Thresholds table To learn what default values are in effect for your system you can produce a dump report for a single two minute interval as follows ADVISE COLLECT REPORT DUMP DATACELLS _ BEGINNING hh mm ENDING hh mm 2 NODE_NAMES nodel hh mm 2 Take the the actual hours and minutes from the beginning statement add two minutes and then enter the final value in the hh mm format Do not use 2 as part of the syntax Look up the values given for their corresponding data cells CPU_VUP_RATING COM_SCALING SOFT_FAULT_SCALING HARD_FAULT_SCALING and IMG_ACT_RATE_SCALING Since the DUMP_DATACELLS report produces voluminous output for each interval and since the processor specific thresholds do not change generate this report for only one two minute interval to learn their fixed values These scaling factors are multiplied by the values shown in the following examples before being applied in rule condition checking m Soft fault scaling factors are multiplied by TD HIGH SOFT FAULT rule literal default value 100 m Hard fault scaling factors are multiplied by TD HIGH HARD FAULT rule literal default value 10 m Image activation scaling factors are multiplied by TD HIGH IMGACT rule literal default value 0 5 Chapter 5 Customize the Knowledge Base 169 Implement Changes m Number in the Compute queue scaling factors are multiplied by TD COM PROCESSES rule literal default value 5
429. nce Agent is a detached process that collects and records performance data for specified nodes in the cluster system according to a weekly schedule Performance evaluation report Periodicity Predefined graph Queue length The performance evaluation report is a statistical report that helps you determine whether changes that you implemented based on recommendations in the Analysis Report improved or degraded system performance Periodicity is a parameter file element for a history file descriptor that specifies how often the averaging cycle is restarted A predefined graph is a graph in which the Performance Manager defines the metrics plotted Queue length is the average number of outstanding requests either waiting for or receiving service Queueing network model Raw data file Recommendation Residence time Response time Queueing network model is a mathematical abstraction of a system where the computer system is represented as a network of queues Each queue in the network is evaluated analytically A recommendation is the text presented in an analysis report that offers system tuning advice based on rules firing See Rule conclusion Residence time is the time in seconds between image activation and image termination Average time that a request spends while waiting for and receiving service Response time is the elapsed time between the arrival of a request and the moment of completion In the context of
430. nd Top Image categories which causes the same graph metric to be displayed Selecting the major category of custom graph updates the lower menu with custom classes Selecting a custom subclass causes the custom graph metrics dialog box to appear For more information about displaying custom graphs see the section Display Custom Graphs see page 295 The arrow buttons allow you to proceed through the categories in the specified class If multiple nodes have been selected for graphing access to graphs by individual nodes is provided Press MB1 on the CLUSTER menu entry to view the popup menu with all selected node names Release MB1 on the node to be graphed The arrow keys allow stepping through the node names sequentially Access to the System graph category is always available by node To access the other graph categories when the data was selected the BY NODE toggle button must have been set Graph format This radio box lets you choose between a graph and a pie chart For graphs displaying data in terms of percentages CPU UTILIZATION Performance Manager provides an additional option If selected the percentage reflected in the pie is 100 with a slice labeled IDLE added to the pie chart Middle section contains 4 buttons arranged vertically This panel contains two toggle buttons for adjusting the graph presentation a lock for locking your settings and a Help button The top toggle button flips the graph from line m
431. nd a Schedule Selection dialog box appears Legend B Collection On O Collection Off Press MB3 inside clock to copy Schedule A 24 hour clock is displayed A bar on the right side of the clock lets you scroll to each day of the week By default data collection is set ON for each hour of the day every day of the week 270 Performance Manager Administrator Guide How You Select Data for Analysis To set the clock 1 Set or reset the square toggle button above the clock to turn data analysis on or off for an entire day To turn off data collection for a specific hour point to the hour on the clock and click MB1 Holding MB1 down and dragging the pointer around the clock will set data collection to the value of the initial hour setting for a series of hours 2 Dragthe slider on the scroll bar to display the collection schedule for each day of the week or click the up or down stepping arrows To duplicate a day s schedule 1 Press and hold MB3 inside the clock A pop up menu is displayed Release on the Cut menu item 2 Scroll to another day and press MB3 inside the clock Release on the appropriate Paste menu item The clock displays the copied schedule The OK button removes the dialog box and applies the new schedule Reset causes the schedule to revert to your previous selection Cancel removes the Schedule Selection dialog box Calendar lets you specify a series of date ranges that are listed in a text file
432. nd closes the dialog Performance Manager allows a panel to be activated or launched from an instrument within a different panel In addition a data item can be passed to the launched panel This data item the instance value can then be used in a subsequent filtering of information displayed in the launched panel s instruments 362 Performance Manager Administrator Guide How You Edit the Panel Instruments To assign launch panels 1 Pull down the Edit menu choose the Build Mode menu item and click the Enable submenu item 2 Select the instrument that will launch a new panel 3 Pull down the Edit menu and choose the Assign Launch Panel menu item Performance Manager displays a Launch Panel Selection dialog as shown in the following screen le File Commands Edit Options CPU Utilization PU Utilization User Processes Launch Panel Selection ee No Launch Panel fres_proftermon J 1 eutfered_vo raze i ero f cPu_modes pecu gg Demonstration When you chose assign launch panel the first metric field in the instrument to be assigned a launch panel will appear with a solid outline If the instrument has more than one field that can launch panels these appear with a dashed outline 4 When you are satisfied with the launch panel assignment for the field and you wish to specify another field in the instrument click Apply and select another field by clicking with MB1 withi
433. nel Manager menu item Performance Manager displays the Real time Panel Manager window as shown in the following screen bAneal time 3 0 0610 L x File Panel instrument Panel Directory D irect 1 0 Disk Memory Allocation Performance Manager displays the Real time Panel Manager window which lets you do the following actions m Specify actions on panels m Close the DECwindows session Specify Actions on Panels To specify actions on selected panels pull down the Panel menu and click the menu item you want The Panel menu lets you Open panels You open a panel either to view it for modifications or to connect it to a node for displaying real time data Click the panel name to select a panel Pull down the Panel menu and click the Open menu item Performance Manager displays the selected panel You can also open a panel by double clicking on the panel name Rename panels Specifies a panel name change 342 Performance Manager Administrator Guide Specify Actions on Panels Create panels Click the Create menu item from the Panel menu and Performance Manager displays the Panel Name dialog as shown in the following screen Type a panel name and click the OK button When you enter a panel name the Panel Name dialog is removed and the name you specified is listed in the Instrument Panel Directory of the Panel Manager Copy panels Click a panel name to select a panel Pull down
434. nents of Pie Charts Each pie chart includes the following components Title For the predefined pie charts the title identifies the type of pie chart Subtitle The pie chart subtitle gives the node name or the list of node names for composite pie charts and the date and time of the selected data Whole Pie Represents The Whole Pie represents the sum of metric values represented by each pie slice Pie Slices The Pie Slices represent the average value of the metric over the time period presented in descending order by value Label The label associated with each slice identifies the metric or item being presented along with the percentage contribution to the pie chart The actual value of the metric can be obtained through either the tabular or CSV pie chart formats Legend The legend identifies the name of the metric and the color pattern or character associated with it Chapter 4 Generate Historical Graphs 115 Pie Chart Presentation of CPU Utilization Pie Chart Presentation of CPU Utilization Pie charts that reflect metrics measured in terms of CPU percentages now have two possible presentations If you specify PERCENTAGE MAXIMUM the pie is drawn in terms of 100 percent with an IDLE section appearing If you specify PERCENTAGE TOTAL the pie is drawn to represent the total of the metrics presented For example if you are producing a pie chart of CPU utilization and the parts of the pie chart have the follow
435. ng displayed CPU percentages are scaled by CPU VUP rating then added Response times I O sizes and disk space are averaged for each node and all other metrics are summed for each node To generate composite graphs specify COMPOSITE on the DCL command line or if in command mode with more than one nodes data selected omit the NODE qualifier When using Windows the CLUSTER option or ALL NODES option generates the composite graph You can recognize a composite graph if there is more than one node listed in the subtitle Note When you produce a composite Graph or Pie Chart and the nodes contribute data to the graph for different time periods possibly missing data the results are undefined Composite Graphs and Pie Charts show accurate totals and averages when the data for the nodes correspond to the same time period Stacked Graphs Graphs showing more than one metric can be displayed either stacked or unstacked When stacked the metric data points are added such that the top most category on the graph plots the sum of all the metrics You can specify that all graphs be stacked with the STACK qualifier or you can specify that none are to be stacked with the NOSTACK qualifier Also each graph you request can have its own specific stack or nostack attribute Stacking is provided by default on some graphs where it makes good sense to do so and the remaining graphs are provided unstacked Specifically rates percentages and cou
436. ng in a CPU limitation 400 Performance Manager Administrator Guide Exit the Character Cell Displays High device CPU usage Look at the CPU mode bars Int interrupt stack A high value for Int might be cause for concern Processes might be blocked from using the CPU because of too many device interrupts Use the ADVISE COLLECT SYSTEM command to collect system wide PC samples and determine the system module usage for example the device driver hence the device s responsible for the excessive interrupts Excessive kernel and or executive CPU time Look at the Ker Kernel and Exe Executive bars in the CPU mode bars If time in Kernel mode is excessive and is not due to page faulting or if time in Executive mode is excessive use the ADVISE COLLECT SYSTEM command to collect system wide PC samples and determine the processes and system modules responsible Interrupt plus kernel CPU time should not be greater than 40 percent of total CPU time Exit the Character Cell Displays To exit the Performance Manager Real time character cell displays m Press Ctrl Z at the PSRT gt prompt Chapter 11 Use the Character Cell Real time Display 401 Appendix A Performance Manager Messages and Recovery Procedures This appendix describes messages that the Performance Manager software generates This section contains the following topics Sample Performance Manager Message see page 403 Severity Codes see page 403 Samp
437. ng with an Analysis Report the performance data for each node follows that node s local analysis The performance data for the cluster appears after the cluster wide analysis Factory analysis reports can be produced only from data collected by the primary data collector namely that which is associated with the CPD definition Interpret the Analysis Reports This section explains excerpts from example Analysis Reports The explanation follows each excerpt The header at the top of each page in the report includes the version number of the Performance Manager used to generate the report This version number does not necessarily correspond to the version number of the Performance Agent that collected the data being analyzed 22 Performance Manager Administrator Guide Analysis Reports Memory Rule Report The following example shows a Memory Rule report Full Analysis NODE01 VAXstation 3100 PA Vx Thursday x 20 FEB 1997 00 00 to 23 59 CONCLUSION 1 While excessive page faulting was occurring there were some users running images which seemed to want more memory than their WSEXTENTS allowed If the WSEXTENTS for these users were larger there may have been less page faulting Increase the WSEXTENT for the following users If batch queue settings for working set extents are causing the limitation increase those If detached processes are causing the problem increase the working set extents specified in either the RUN
438. ning a scale set at the current hotfile reporting count as shown in the following screen Number of files to report per device Cab Drag the arrow to change the number of files that you want reported for each disk device Control Buttons m The OK button applies the change and removes the dialog box m The Reset button restores the value to your last selection if you previously changed the default or to the default value of 20 m The Cancel button removes the dialog box without changing the limit Chapter 8 Use the DECwindows Motif Interface 281 How You Display Analyzed Data Process Statistics Release on the Process Statistics menu item from the Display menu to open the Performance Manager Process Statistics Window as shown in the following screen X Process Statistics File Customize Help Traditional Process Statistics Start Time 26 JUN 2006 10 36 End Time 26 JUN 2006 11 36 Explanation The table below lists observed workload characteristics of all the interactive images that were run during the given interval Note that Diskio Bufio and Cputim are percentage contributions of the respective images to the total system load Working set size and Node Name I64VM6 MODE INTERACTIVE of Page Faults Avg of of Uptime Cputim activ per Actvtn Ws Direct Buffered X of image image Image ations Soft Hard size 1 0 I 0 Cputim sec sec dcl 0 123 0 197 0 39 1 77 0 00 1464 0 17 ANAL YZO
439. ns A rule condition is an expression that the Performance Manager resolves to true or false status during the course of rule evaluation A rule condition must end with a semicolon If the Performance Manager evaluates the rule condition to 1 0 then the rule condition is true otherwise it is false A rule condition is composed of rule expressions A rule expression is one of the following values m decimal value m literal symbol m tally data cell m rule expression m numeric data cell index specifier data cell boolean data cell index specifier data cell m scan routine data cell rule expression m string item string operator string item m rule expression numeric binary operator rule expression 152 Performance Manager Administrator Guide Rule Expression Operators Components of Rules and Descriptions A numeric or string operator is one of the symbols listed in the following table Symbol Definition EQS String equal to string NES String not equal to string i Multiply Divided by Plus x Minus EQ Equal to NE Not equal to LT Less than EE Less than or equal to GT Greater than GE Greater than or equal to AND And OR Or The Boolean operators AND and OR evaluate to either a decimal 1 0 for true or a O for false The operands are treated as true if their values equal decimal 1 0 and anything else as false Valid components of rule expressions are defined as
440. ns COM CFG CPU DSK FIL LOC PRO TAP IMAGE NAMEPRO A IMAGENAME This contains a string indicating the image name for which the current process subrecord pertains on the local node for the current interval Data Type STRING Domains PRO IMAGE TERMINATION RATEMET F IMGTRMS This contains the number of image terminations per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IMG_ACTIVATIONS_PER_PID Derived This contains a value representing the average image activation rate for the PID indicated by the process subrecord indexed by the cell HIGH IMAGE ACTIVATIONS PID X on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP IMG_ACT_RATE_SCALING Derived This contains a value representing scaling factor for the image activation rate for the local node The value is obtained from the threshold TD_IMG_ACT_SCALING_n where n is the hardware model number of the local node By default if the local node is a VAX 11 780 the value would be 1 0 The value of this data cell can be modified using a threshold construct Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 457 Performance Manager Data Cells INCOMING BLOCKING AST RATEMET F BLK IN This contains the number of incoming blocking ASTs queued per second for the current interval for the local node Data Type
441. nst which Performance Manager can report Use ADVISE EDIT to create modify and delete workload definitions A Transaction workload definition contains at least one image name Typically this workload will contain images with similar resource demands A Usergroup workload definition contains a user specifier such as accountname username processname UIC code Typically this workload will contain a group of users who belong to the same business unit such as a department Transaction and Usergroup are terms applied to a workload by the CLASSIFY BY qualifier A workload family is a collection of workload definitions collectively representing the entire work on a computer system Workload family name Workload name A workload family name is a name that identifies a group of workloads that collectively constitute a unit called a family A workload name is a name that identifies a workload or workload definition Glossary 575 Workload selection criteria A workload selection criteria is the criteria by which process data is assigned to a workload A workload is selected when process data matches on either or both of a user specifier or image name 576 Performance Manager Administrator Guide Index A ADVISE PERFORMANCE command e 175 for building a knowledge base e 175 177 for converting data e 182 for graph generation e 191 230 245 for Performance Manager version e 238 for pie chart generation e 232 249 for rep
442. nt 1 Pull down the Edit menu choose the Build Mode option and click the Enable submenu option 2 Pull down the Edit menu choose the Create menu item and click the submenu entry Chapter 10 Customize the DECwindows Motif Real time Display 357 How You Edit the Panel Instruments Copy Instruments Delete Instruments You can choose from the following instruments m Strip Chart Bar Graph m Meter To copy instruments 1 Pull down the Edit menu choose the Build Mode option and release on the Enable submenu option 2 Select the instrument 3 Pull down the Edit menu and choose the Copy menu item The cursor changes to an indicator that represents the upper left corner of the instrument Position the cursor at the desired location for the new instrument in the panel Click MB1 and the instrument is displayed To delete instruments 1 Pull down the Edit menu choose the Build Mode menu item and release on e Enable submenu item 2 Select the instrument 3 Pull down the Edit menu and choose the Delete menu item Performance Manager displays the following message box Li Do you really want to delete the instrument 358 Performance Manager Administrator Guide Assign Metrics How You Edit the Panel Instruments To specify metrics to be displayed 1 Pull down the Edit menu choose the Build Mode menu item and click the Enable submenu item 2 Select the instrument 3 Pull down the Edit menu an
443. nt of resources such as a database update Changes in the affect the residence time of jobs that use a fixed amount of resources In the CPU Utilization histogram interrupts designated by X used approximately 5 percent of the CPU interactive jobs used 10 percent of the CPU and batch jobs used 85 percent of CPU at approximately 11 30 a m on October 1 2007 An asterisk appears in the histogram if there is a discrepancy between the total CPU utilization and the utilization accounted for by processes This can happen if the image activation rate is high and the Performance Agent cannot capture all of the image activity 64 Performance Manager Administrator Guide Histograms CPU Utilization Histogram The following example shows a CPU Utilization histogram Histograms DEMAND VAX 8700 Wednesday 01 OCT 2007 00 00 to 12 16 Legend 1 DECnet jobs interactive batch overhead swapper netacp intrupt amp mpsync CPU utilization BBB BBB BBB BBB BB BBB BBB BBB BBBB BBB BBBB BBB BBBB BBB BBBB BBB BBBB BBB BBBB BBBB BBB II B BBBB IBBB BB BBBB II BB BBBI IBBBBBB BBBB II BB BBBII IBBBBBB BBBB I II BBBBBBII IBBIBBB BBBB IIIIIBIIIIBBIII IBIIBBB B BBBB I IIIIIIIIIIBIIII IIIIIIIB B BBBB IIIII IIIIIIIIIIIIIIII XXXXXXXB B B B B B XIBXI IXXXXXXXXXXXXXXXXXXXXXXXX 01 02 03 04 05 06 07 08 09 10 11 12 3 Each Column represents approximately 10 minutes starting from 1
444. nt to perform It reflects the status of your use of the application and if available the status of data collection in your environment b4 Unicenter Performance Management by CA lo Control Select Display Customize View Help PSDC DATABASE translation System dfn Collection Status SYS COMMON DECPS DATABASE System dfn Reporting Customizing SYS COMMON DECPS DATABASE Aa Agent Status l64Box Running Data Selected for Processing No data selected Files locked by this session No files locked KL The Performance Manager Main Window lets you do the following tasks m Control the DECwindows interface m Select data for analysis m Display analyzed data m Customize data collection PSDC DATABASE definition and the parameters file m View or remove specified main window sections m Get Help either Contextual specific to a widget or General relating to a window 258 Performance Manager Administrator Guide Start the DECwindows Motif Interface Main Window Status Information The Performance Manager Main Window displays the following information PSDCSDATABASE Translation Displays the following directory information Collection Status Displays the system wide definition of the logical name PSDCSDATABASE If there is a data collector running on your analysis node this directory contains the data files produced by the Performance Manager and the schedule file for controlling the collection Repor
445. nterval represented by this record Data Type NUMERIC Domains PRO PROCESS COM PERCENTPRO F COMPU This contains a value representing the percent of time in the computable state over the processes uptime for the current process subrecord on the local node for the current interval Sampled every 5 seconds Data Type NUMERIC Domains PRO Appendix C Performance Manager Data Cells 481 Performance Manager Data Cells PROCESS COM PERCENT TALLY Derived This contains the sum of the values representing the percent of time in the computable state for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS CPUTIMEPRO F CPUTIM This contains a value representing the CPU time in milliseconds for the current process subrecord on the local node Data Type NUMERIC Domains PRO PROCESS CPUTIME TALLY Derived This contains the sum of the values representing the amount of CPU time of all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS_CURRENT_PRIORITYPRO_B_PRIB This contains a value representing the process s current priority This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Do
446. nting the file specification of the cache for the current disk on the local node for the current interval Data Type STRING Domains DSK DISK_CONTROLLERDEV_A_CTLR_NAME This contains a string indicating the controller name of the current disk on the local node for the current interval e g DUA Data Type STRING Domains DSK DISK_DINDX_CACHE_SIZEDEV_F_DINDXSIZE This contains a value representing the number of entries in the directory index cache for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK 434 Performance Manager Administrator Guide Performance Manager Data Cells DISK DIRDATA CACHE SIZEDEV F DIRSIZE This contains a value representing the number of entries in the directory data cache for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK ERROR COUNTDEV F ERRCNT This contains a value representing the number of errors recorded for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK ERROR COUNT TALLY Derived This contains the sum of the values representing the number of errors for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_EXTENT_CACHE_SIZEDEV_F_EXTSIZE This contains a value representing the number of entries in the file e
447. nts are stacked by default and response times I O size and disk space utilization are unstacked by default Create Typical Time Period Graphs If you want your graphs to represent a typical time period such as an average day you can turn on graph averaging On the DCL command line specify AVERAGE DAILY WEEKLY MONTHLY and QUARTERLY are also options From command mode the AVERAGE qualifier must be specified with the SELECT verb Window users can set this option from the Select Data dialog box under Additional Options Chapter 4 Generate Historical Graphs 113 Scheduling Scheduling If you want a typical Monday type graph you specify AVERAGE WEEKLY and specify a schedule with only Monday as indicated in the following example SCHEDULE NOEVERYDAY MON 0 24 If history data with the periodicity attribute set is used for the graph the history s periodicity is used for and overrides the graph averaging If you want certain time periods included and others not included on the graph you can use a combination of the schedule and dates features The SCHEDULE qualifier and the schedule time clocks in the window interface allow you to specify desired hours on a weekly basis for inclusion on the graph If you need more specific selection time frames you can use a DATES file see the DATES qualifier which specifies an unlimited number of date ranges to indicate the desired time frames for inclusion in the graph
448. nvestigate Displays LOAD BALANCE Display ANSI zi DE Cterm 1 File Edit Commands Options Print AE GALLO I5 Ftti 1992 16 12 Investigate a Memory Limitation The chief indicators of a memory limitation are as follows No free memory Look at the value of free pages the memory bar histogram and the memory queue Mem que If the value of free pages approaches that of FREELIM the system parameter that sets a lower limit on the number of pages on the free list this indicates a memory shortage The memory bar histogram shows the amount of memory used by the free and modified page lists with respect to that available for user working sets Again a shortage of memory for the page caches free plus modify lists indicates a high degree of user memory utilization and consequently a memory shortage A nonzero value of Mem que indicates processes in the memory queue in computable outswapped states awaiting memory that is unavailable Page fault high Look at the value of Pgflts page faults Sysflts system faults and paging in the IO window for disks 394 Performance Manager Administrator Guide Evaluate Performance Using the Investigate Displays For a VAX 11 780 CPU a value of Pgflts page fault rate or number of faults hard and soft including system per second greater than 100 is cause for concern For other CPUs use an appropriate threshold such as that supplied with the factory knowledge base If Pgflts is
449. o of time in user mode 127 7 0 0 compat 9 o of time in 136 7 compatibility mode 40 0 idle of time in idle mode 145 7 58 0 system cpu of time in system 154 7 mode I K E 1 3 task cpu 9o of time in task mode 163 7 S U C 2315 0 extcpu_samp number of extended CPU 172 7 les samples 34 4 cpu_busy of time CPU found 181 7 busy 0 1 swap_busy of time swapper busy 190 7 4 8 multio busy of time more than 1 199 7 disk busy Appendix E Output Format for ASCII CSV Data 543 Secondary CPU Statistics Data Record Example Data item 35 1 anyio busy 9 8 pagewait 0 0 swapwait 9 8 mmgwait 42 9 sysidle 22 0 cpu only 12 4 cpu io 22 7 io only 0 com state 1 cpu id 0 3 busy wait Description of time when at least 1 disk busy of idle time with page i o outstand 9 o of idle time with swap i o outstand Yo of idle time with page or swap i o outstanding of time CPU and disks idle 9 o of time CPU busy and disks idle 9 o of time CPU busy and at least 1 disk busy of time at least 1 disk busy and CPU idle number of processes in computable state CPU id number e g BI node number CPU board s of time in busy wait spin time Position 208 217 226 235 244 253 262 272 280 289 294 Length 7 Secondary CPU Statistics Data Record The following example output shows these Secondary CPU statistics ULTRA VERS Vx x REPORT LOG FILE Vx x CPD
450. o process x nes BACKUP Image name Top Dirio process x nes VMSBUXX Occurrences td XQP cache Chapter 5 Customize the Knowledge Base 165 Implement Changes Add a New Rule Evidence Sysgen Acp Hdrcache Disk Header Cache size Highest IO rate disk x File header Cache HR File header Cache AR Volume name top dskio proc topdsk x Disk IO rate top dskio proc topdsk x User name top dskio process x Image name top dskio process x Time Conclusion There are too many disk I Os caused by a low hit ratio on the file header cache This will occur if your workload causes disk files to be scanned instead of repeatedly accessed i e BACKUP DIR SEARCH etc However if your workload does not scan disk files so that there is still useful information in the cache then you may benefit by using AUTOGEN with the feedback mechanism to automatically increase the SYSGEN parameter ACP HDRCACHE After successive uses of AUTOGEN its feedback mechanism provides the system with sufficient file header cache for the average workload Brief conclusion Low hit ratio high attempt rate on the file header cache EndRule Assume that a user would like you to add a rule that fires when any disk is low on space The following data cells DISK MAX BLOCKS ANY DISK FULL a binary data cell type derived from DISK MAX BLOCKS and DISK MOST FULL X an index specifier pointing to the most affected disk make this type of rule possible
451. o specify a list of processes in the Collection Coverage List box You specify a list of processes to be included in collection or excluded from collection All devices When you select the collection of IO data the All Devices option is collected by default If you have modified the device list the menu is revised to reflect the type of coverage list include or exclude and the count of specified devices Clicking on this menu will display the correct list This menu provides a convenient way to view the coverage list Modify device list Click this button to specify a list of devices in the Collection Coverage List box You specify a list of devices to be included in collection or excluded from collection 302 Performance Manager Administrator Guide How You Customize Coverage Lists For Process and Disk classes you can collect data for specific lists of items or exclude lists of items Use option menu to select coverage All processes o To create a list of processes push down and hold on the uppermost box in this window which displays All processes or one of the other options When the menu appears pull down to the menu item you want and release Click the text entry field to activate the cursor To add a process name enter a name and press Retum or click the Enter button The name will appear in the list box and be cleared from the text entry field To delete a name click the name in the name list box
452. ode to fill mode and back again The pixmap on the button reverses to be the opposite of the state of the graph allowing you to see where you can go Chapter 8 Use the DECwindows Motif Interface 291 How You Display Analyzed Data The second toggle button changes the graph presentation from stacked to unstacked and back again A stacked graph displays the metrics one on top of another An unstacked graph displays each metric value with respect to the X axis This can cause occlusion of some details so it is best to view unstacked graphs in line mode The lock button will lock your current settings for stack unstack line fill mode Y axis maximum and threshold line selection As you view different graphs the same settings will be applied m Y Axis maximum Specifies a new value for Y Axis Maximum This lets you minimize the affect of spikes on graph presentation and to gain consistency in graphs generated from different data selections You can either drag the pointer across the scale or click either side of the value to adjust the scale Press and hold MB1 to scroll the values The Y Axis Maximum value is maintained until the next data selection m Threshold line Specifies a line to be displayed on the graph reflecting the value of the selected button You can also specify a specific value at which a threshold line is to be displayed The Max Min and Ave lines are available only for stacked graphs Change Color Releasing MB1
453. of the intervals Data Type NUMERIC Domains SUMMARY AVERAGE WORKING SET SIZE Derived This contains the value of the average working set size for all processes for the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP AVG NONPAGEDPOOLBYTES INUSE Derived This contains a value representing the average number of non page pool bytes in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY AWSA IS SLOW Derived This contains a Boolean value zero or one where one truth represents the presence of slow automatic working set adjustment for 2 or more processes on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BADPAGE FAULT RATEMET F BADPAGE FAULTS Average number of page faults per second from the bad page list for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 415 Performance Manager Data Cells BATCH COUNTMET F BATCH This contains a value representing the average number of batch processes on the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP BIG WS AND BIG QUOTAS Derived This contains a Boolean value representing the presence of large working sets and quotas on the local node for the current interval Data Type NUMERIC
454. okups second 100 0 fid_hit percent file id entries 127 7 found in cache 0 2 fid_rate file id cache 136 7 lookups second 100 0 extent_hit percent extent entries 145 7 found in cache 0 3 extent_rate extent cache 154 7 lookups second 80 8 filhdr_hit file header entries 163 7 found in cache 5 5 filhdr_rate file header cache 172 7 lookups second 88 7 dirdata_hit directory data entries 181 7 found in cache 4 5 dirdata_rate directory data cache 190 7 548 Performance Manager Administrator Guide lookups second Example 0 0 0 0 System Communication Services Data Record Data item Description Position Length stormap hit storage bitmap entries 199 7 found in cache stormap rate storage bitmap cache 208 7 lookups second System Communication Services Data Record The following example commands show how to export and display system communication statistics ADVISE PERFORMANCE EXPORT NODE SNOLPD OUTPUT EXP SYS _ CLASS NODEFAULT SYSTEM COMMUNICATION _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP SYS SNOLPD VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 SNOLPD SYST 120 2 SNOLPD 0 0 0 0 0 0 10 2 0 0 0 0 4 5 4 5 0 0 0 0 0 0 0 0 Example SNOLPD 0 0 0 0 0 0 10 2 0 0 Data item Description Position Length scs node name of node 92 17 data gs recvd data gram msg 111 7 rece
455. ol 5 Number of hours spanned by the reporting period This histogram reflects a reporting period of 24 hours 70 Performance Manager Administrator Guide Tabular Report Sections Tabular Report Sections The Tabular Report provides statistics summarized by classes of metrics These classes include CPU DISK IO LOCK MEMORY PAGING PROCESS SCS and CACHE The report classes are accessed by using the SECTION qualifier The sections include the following classes m Configuration Section Overview section listing node collection interval and reporting interval information m Summary Section Presents classes CPU IO LOCK MEMORY PAGING SCS and CACHE m Disk Section Presents the DISK metrics class m Process Section and Extended Process Section Presents the PROCESS metrics class Because the Tabular Process Metrics displays the information by PID IMAGENAME PROCESSNAME ACCOUNT and USERNAME you must specify all these key levels when selecting data in either command mode or via the DECwindows interface The Tabular report can be requested in either a final form which presents the data summarized over the entire reporting period specified or in an interval form which presents a series of sub reports for the reporting interval specified For example if the overall time period indicated by the BEGIN and END qualifiers is one hour and the INTERVAL qualifier is used with a value of 600 seconds each requested report se
456. ollowing font characteristics Family Size Weight Slant Font options not available on your server will be disabled Chapter 10 Customize the DECwindows Motif Real time Display 355 How You Edit the Panel Instruments Data Tags Specifies whether the tags are displayed on the left or right side of the chart or both Tick Marks Specifies whether tick marks are displayed on the left or right side of the chart or both Tick marks can be displayed without tags When tags are selected tick marks are displayed automatically Number of Tags Tick Marks Specifies the number of tags and corresponding tick marks to be displayed The default is six tags Number of Short Tick Marks Specifies the number of minor tick marks to be displayed The default is one Units Label Specifies whether the units label is included in the display Time Display Specifies whether the time display will include hours minutes and seconds or a subset of these Tick Marks Specifies whether tick marks are displayed on the top or bottom of the chart or both Units Labe Specifies whether the time units label in included in the display Labels Font The Font Selection dialog lets you specify font characteristics Meter Parts The Meter Parts Definition dialog lets you specify the following attributes Title String Set the toggle button to display a title Enter a title in the text entry box Data Name Set the toggle button to specif
457. om 6 SEP 00 00 00 to 6 SEP 23 59 00 An N indicates NO DATA The following statements are keyed to the columns in the previous example 1 Explanatory list of units in the histogram columns 2 Percentage of memory used 3 Number of hours spanned by the reporting period This histogram reflects a reporting period of 24 hours 66 Performance Manager Administrator Guide Histograms Disk I O Per Second Histogram The following example shows a disk I O per second histogram Histograms DEMAND VAX 8700 Page 7 PA Vx x Tuesday 06 SEP 2005 00 00 to 23 59 Legend 1 user io DISK I 0 PER SECOND P pag swping per sec 2 kk kx kk k kk Seekekpphoek ck kk kx Kk oook p poelek poe pok akakakakok px 444pp ppppp pppppppxpppp x xxx Xx p perppppseckpespp ppp ppPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPP P pp pppp 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 3 Each Column represents approximately 20 minutes starting from 6 SEP 00 00 00 to 6 SEP 23 59 00 An N indicates NO DATA The following statements are keyed to the previous example 1 Explanatory list of units in the histogram columns 2 The number of disk I Os per second attributable to either user I O or paging and swapping 3 Number of hours spanned by the reporting period This histogram reflects a reporting period of 24 hours Chapter 3 Evaluate Performance in Detail 67 Histograms Terminal I O Per Second Histog
458. omains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN AWSMINPAR F AWSMIN The value of the SYSGEN parameter AWSMIN which establishes the lowest number of pages pagelets to which a working set limit can be decreased by automatic working set adjustment for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN AWSTIMEPAR F AWSTIME The value of the SYSGEN parameter AWSTIME which specifies the minimum amount of processor time that must elapse for the system to collect a significant sample of a working set s page fault rate for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 504 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN BALSETCNTPAR F BALSETCNT The value of the SYSGEN parameter BALSETCNT which is the number of working sets which determines the maximum number of processes that can be concurrently resident for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN BORROWLIMPAR F BORROWLIM The value of the SYSGEN parameter BORROWLIM which defines the minimum number of pages required on the free page list before the system
459. omponents of Graphs see page 111 Composite Graphs see page 113 Stacked Graphs see page 113 Create Typical Time Period Graphs see page 113 Scheduling see page 114 Use Binary Graph Data see page 114 Components of Pie Charts see page 115 Pie Chart Presentation of CPU Utilization see page 116 Format Graphs and Pie Charts see page 116 Generate Custom Graphs see page 126 Graph the Hot File Activity see page 133 Generate Predefined Graphs You can generate predefined graphs in the following three ways m At the DCL level see the section Generate Graphs from the DCL Level see page 108 m Within the Performance Manager command mode see the section Generate Graphs in Command Mode see page 108 m Within the DECwindows interface see the chapter Use the DECwindows Motif Interface see page 257 Regardless of the method you use the basic process is the same The Performance Manager selects reads and buckets the data then formats and writes the graph to the output device Chapter 4 Generate Historical Graphs 107 Generate Predefined Graphs Each method has its own advantages and disadvantages in terms of ease of use efficiency and equipment This section discusses how to create single and multiple graphs from the DCL level and within the Performance Manager command mode The DECwindows Motif Interface is discussed in the chapter Use the DECwindows Motif Interface see page 257 Generate Gra
460. omponents of Rules Name Default Description Value TD T56 RF31 78 Same as previous TD T57 RF72 37 Same as previous TD T58 RAM DISK 130 Same as previous TD T59 RZ25 38 Same as previous TD T60 RZ56 34 Same as previous TD T61 RZ57 38 Same as previous TD T62 RX23S 10 Same as previous TD T63 RX33S 10 Same as previous TD T64 RA92 33 Same as previous TD T65 SSTRIPE 60 Same as previous TD T66 RZ23L 30 Same as previous TD T67 RX26 10 Same as previous TD T68 RZ57I 38 Same as previous TD T170 RZ58 4 Same as previous TD T171 SCSI MO 20 Same as previous TD T72 RRD42 20 Same as previous TD 173 CD LOADER 1 20 Same as previous TD T74 ESE25 240 Same as previous TD 175 RFH31 78 Same as previous TD 176 RFH72 37 Same as previous TD T177 RF73 44 Same as previous TD T178 RFH73 44 Same as previous TD T179 RA72 40 Same as previous TD T80 RA71 40 Same as previous TD T81 RF35 69 Same as previous TD T82 RFH35 69 Same as previous TD T83 RF31F 40 Same as previous TD T84 RZ72 37 Same as previous TD T85 RZ73 36 Same as previous TD T86 RZ35 53 Same as previous Chapter 5 Customize the Knowledge Base 143 Components of Rules Name Default Description Value TD T87 RZ24L 35 Same as previous TD T88 RZ25L 41 Same as previous TD T89 RZ55L 34 Same as previous TD T90 RZ56L 37 Same as previous TD T91 RZ57L 42 Same as previous TD T192 RA73 44 Same as previous TD T93 RZ26 50 Same as previous TD T94 RZ36 51 Same as previous
461. on Change a Rule Value see page 172 164 Performance Manager Administrator Guide Implement Changes The following example shows how your auxiliary rules file might look The bulk of the new rule came from the corresponding factory rule in PSPASEXAMPLE PSPASKB VPR The only difference is the addition of a condition to filter out VMSBUXX image records from the rules processing The actual rule number could not simply be copied in the auxiliary rules file as a zero in the second character is reserved Most of the conclusion sections were copied from their corresponding sections in SYS SYSTEM PSPA MSG TXT The example also shows that you can make minor modifications to the factory rules without investing significant time and effort To use this rule file see the sections Build an Auxiliary Knowledge Base see page 172 and Use an Auxiliary Knowledge Base for Reporting and Archiving sections see page 173 To change an Existing Rule m Use the following code DISABLE R0095 Literal TD FILE CACHE HITRATIO 70 TD FILE CACHE MISSEDIO RATE 5 TD XQP CACHE 10 EndLiteral Rule UR095 XQP Cache hit ratio lt td file cache hitratio XQP Cache missedio rate ge td file cache missedio rate File header Cache HR lt td file cache hitratio 100 File header Cache HR 100 File header Cache AR ge td file cache missedio rate Disk header cache size Highest IO rate disk x ge Sysgen Acp Hdrcache Image name Top Diri
462. on Virtual QIO Interpret cluster wide Summary Statistics In the Performance Evaluation Report the summary statistics follow the cluster wide hot file statistics The summary statistics highlight the CPU and memory utilization for the configuration To display only the summary statistics section of the report use the following qualifier INCLUDE SUMMARY STATISTICS To disable the summary statistics display from the report use this qualifier INCLUDE NOSUMMARY STATISTICS 62 Performance Manager Administrator Guide Performance Evaluation Performance Evaluation Report The following example shows the summary statistics section of the Performance Evaluation report CLUSTER Page 1 PA Vx x Tuesday 26 JAN 2006 09 00 to 10 00 The following table gives a summary of the average CPU and MEMORY utilization and average number of jobs by type for each node LATOUR VAX 8700 YQUEM VAX 6000 440 GALLO VAX 8700 sUtil Intractv Batch Netwrk Detach 50 3 35 38 0 16 4 13 40 02 62 6 115 05 3 00 6 34 41 00 85 1 30 66 1 00 4 48 56 01 User Command ADVISE PERF REPORT PERF BEG 26 JAN 2006 09 00 00 00 END 26 JAN 2006 10 00 00 00 INCLUDE SUMMARY QUT EXAMPLE 3 13 INFO 9 The following statements are keyed to the previous example 1 2 so coo M DO SU Node name Type of processor LATOUR is a VAX 8700 YQUEM is a VAX 6000 400 and GALLO is a VAX 8700 Average percentage of time that each node s CPU was used d
463. on an option in the Change Color submenu causes Performance Manager to display a color mixing dialog box The colors you establish are used when you save a color PostScript graph By Node Releasing MB1 on a node name in the By Node submenu causes the current graph and all subsequent graphs to be drawn for that node only Note If you did not select the By Selected Node processing option in the Data Selection dialog box only system metrics graphs are available by node All Nodes Releasing MB1 on All Nodes causes the current graph and all subsequent graphs to be redrawn for all nodes selected When all nodes are selected the graph becomes a Composite graph See the chapter Generate Graphs see page 191 for more information on Composite graphs 292 Performance Manager Administrator Guide How You Display Analyzed Data Display Predefined Graphs Pull down the Display menu to view the possible system metrics graphs Release on the type of graph you want Your choice is displayed in the Performance Manager Graph Window as shown in the following screen File Edit Display Display Top Display Custom COMPU gt PROCE Chapter 8 Use the DECwindows Motif Interface 293 How You Display Analyzed Data Display Top System Use Graphs Pull down the Display Top menu to display graphs of top system use Point to a Top menu item and click the submenu icon Click a menu item in the submenu The graph you have chosen is displayed as shown in
464. only way to recover the selected data is to reselect A faster way is to SAVE the selected data so that if later review of the same time period is required you can use the load function for much faster retrieval of the data Save the Binary Graph Data To save the binary graph data 1 Release on the Save Binary Graph Data menu item Performance Manager displays the Save Graph Data Dialog Box Data Dialog Box Save Graph Data in binary loadable format as 1 a cancel Help 2 Specify a file to which the currently selected graph data is to be saved in a binary format All graph data points are saved and the file can be reloaded using the Load Graph Data Dialog Box The save operation does not affect the current selection You can also load and save this binary graph data using the DCL command mode interface Chapter 8 Use the DECwindows Motif Interface 265 How You Select Data for Analysis Quit the Session To end a DECwindows session 1 Click the Quit menu item If you have any outstanding changes to the Schedule file or the Parameters file you are asked if you want to save them 2 Click Yes to save your changes OR Click No to quit without saving your changes The interface proceeds to end of job How You Select Data for Analysis The DECwindows Interface lets you analyze performance data and display graphs pie charts and reports including the Analysis Report Performance Evaluation Report
465. ons User Memory Panel Page File Allocation Panel Label Metric Units Next Panel Total Memory Memory utilization for this Pages Pages for User user CPU Utilization CPU utilization for this Percentage user Disk IO Rate Disk I O for this user Operations per second Disk Thruput Disk throughput Kilobytes per second Hard Fault Rate Hard page fault rate Faults per second Soft Faults Rate Soft page fault rate Faults per second Buf I O Rate Buffered I O rate Operations per second Direct IO Rate Direct I O rate Operations per second User Processes Memory utilization by Percentage list process for this user Label Metric Units Next Panel Page File Write Page file write rate Pages per Rate second Page Swap I O Page Swap I O Rate by Faults per Rate Top volume second Volumes list Read Rate Write Rate Read I O Rate Write I O Rate Page read operation rate Page write operation rate Page read disk operation rate Page write disk operation rate 334 Performance Manager Administrator Guide Operations per second Operations per second Operations per second Operations per second Default Panel Descriptions Label Metric Units Next Panel Swap Busy Swapper busy Percentage Swap Wait Swapper wait Percentage Disk Rate Panel Descriptions Disks Panel Volume Info Panel The Disk Rate graph on the System Overview panel lets you access the following pane
466. or A leading exclamation point makes a setting a comment Only override those items that are significant to you as the process does slow the activation of the Data Selection dialog box Chapter 8 Use the DECwindows Motif Interface 277 How You Display Analyzed Data How You Display Analyzed Data To display reports or graphs pull down the Display Menu and release on the option you want The Display menu lets you generate the following reports m Brief Analysis Reports Full Analysis Reports m Performance Evaluation Reports m Process Statistics Reports m Tabular Reports m Graphs and pie charts Dump Reports For more information about dump reports see the Performance Agent Administrator Guide Until a complete data selection process has occurred these options are unavailable Brief Analysis Report Release on the Brief Analysis menu item from the Display menu to open a Performance Manager Analysis Report Window Regs Start Tane 26 JAN 193 COD End Tine 26 ANN 193 10400 wa TOUR Summary for LATOU O rules fired of 30 records 2 satisfied conditions Line Description leed balance anafor 400 disks eetielied condi tiene The Brief Analysis Report lists the rules that fired during the analysis period Click anywhere on the rule and Performance Manager DECwindows opens another Analysis Report Window that displays the fired rule s conclusions conditions and evidence To return to the Brief Analysis Rep
467. or Guide request Average number of milliseconds to process an I O operation Note that this is zero if there are no I O operations Number of split I O operations per second to the disk Number of free blocks on the disk Number of MSCP I O operations per second Number of paging and swapping I O operations per second Percentage of free disk space for a given disk Percentage of used disk space for a given disk Channel QUEUE READIO THRUPUT TOTIO WRITIO Disk User USER VOLUME IO HSC HSC DISK IO HSC DISK THRUPUT HSC IO HSC TAPE IO HSC TAPE THRUPUT HSC THRUPUT File Metric FILE TOTIO FILE PAGIO FILE READIO ADVISE PERFORMANCE GRAPH Description Average number of I O operations outstanding Number of read I O operations per second Number of Kbytes per second transferred to or from the disk Number of I O operations per second Number of write I O operations per second Description Number of I Os per second for the user s use of the disk volume This is based on the collected top two disks I O rates per process Description Number of disk I O operations performed by the HSC Number of bytes per second transferred to and from disks on the HSC Number of I O operations transferred by the HSC Number of tape I O operations performed by the HSC Number of bytes per second transferred to and from tapes on the HSC Number of bytes per second transferred by th
468. or composite graphs the date and time of the selected data the number of x axis data points and the time represented by each point Axis Labels Time is implied as the x axis label Labels on the y axis specify the units of the plotted values for example Percent of CPU X and Y Axis Markers Axis markers indicate the magnitude and time of any point on the graph The x markers indicate the time The y markers are scaled based on the maximum value of all the data points 286 Performance Manager Administrator Guide How You Display Analyzed Data Legend The legend appears at the bottom of the graph The legend identifies the name of the metric and the color or pattern associated with it The Performance Manager Graph Window lets you do the following Save a Graph Save the graph Edit the graph format Display predefined graphs Display Top system use graphs Display Custom graphs To save the graph or pie chart 1 Pull down the File menu and release on the Save as menu item The Performance Manager Save Graph or Save Pie dialog box is displayed as shown in the following screen Save Graph Dialog Box Graph Title CPU UTILIZATION Select Output Format Ansi KmNheight 24 Tabular Ansi mNheight 60 csv PostScript Color ReGIS Color Q PostScript B W gt ReGIS Bow Save Pie Dialog Box Pie Chart Title ev UTILIZATION Select Output Format 4p PostScript Color
469. or historical data The following section describes the qualifiers you can use with the ADVISE PERFORMANCE REPORT command to control report generation The report keywords are as follows m ANALYSIS m Consists of conclusions conditions and evidence for each rule that fired for each node and includes cluster wide conclusions for a cluster system m BRIEF ANALYSIS A brief version of the analysis report consisting of a one line synopsis of each conclusion m HISTOGRAMS Consists of chronological charts that show peak resource usage The Performance Manager produces a report containing separate histograms for CPU utilization number of disk I Os number of terminal I Os memory usage node status information and if you include the IMAGE qualifier an image residence time histogram Use of the IMAGE qualifier it limited to the DCL command interface m PERFORMANCE EVALUATION m TABULAR FINAL INTERVAL BYCLUSTER BYNODE Contains an overview of the system activity on a per node basis or cluster wide Subsections of this report can be selected or omitted with the SECTION qualifier The Tabular report can be presented in the following ways FINAL Each Tabular report section is presented with statistics representing the whole time period INTERVAL Each Tabular report section is presented for each reporting interval By default the reporting interval is the same as the recording interval however you can specify the reporting
470. or it to fire The occurrence element ends with a semicolon When a rule fires an entry is made in the Analysis Report You can specify evidence for every rule You can select one or more data cells to use in the evidence list You cannot use scan routine data cells as evidence When the rule fires all rule conditions are true and the occurrence threshold is met the Performance Manager saves the current values for all data cells listed as evidence The format for an evidence element is as follows EVIDENCE evidence value evidence value Evidence value is a data cell that is not a scan routine data cell index specifier index specifier The evidence element ends with a semicolon When a rule fires an entry is made in the Analysis Report The Performance Manager starts preparing conclusions if any after all data has been processed After the conclusion the Performance Manager lists the evidence in the Analysis Report in tabular format Each column corresponds to an evidence data cell and is indicated by an appropriate column header 154 Performance Manager Administrator Guide Components of Rules If there are many evidence fields the width of the evidence columns could exceed 80 characters If when writing user auxiliary rules you do not want the width to exceed 80 characters then you must limit the number of evidence items The width of each evidence column depends on the column header and the width of
471. or the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 502 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN ACP EXTCACHEPAR F ACP EXTCACHE The value of the SYSGEN parameter ACP EXTCACHE which sets the number of entries in the extent cache for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN ACP EXTLIMITPAR F ACP EXTLIMIT The value of the SYSGEN parameter ACP EXTLIMIT which specifies the maximum amount of free space to which the extent cache can point expressed in thousandths of the currently available free blocks on the disk for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN ACP FIDCACHEPAR F ACP FIDCACHE The value of the SYSGEN parameter ACP FIDCACHE which sets the number of file identification slots cached for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN ACP HDRCAC
472. ore nodes click the entries in the list and then click Delete to the left of the box To change the order of the nodes in the list click the node to be moved and then the arrows in the right hand corner Only one node can be moved at a time The changes you make in the node definition section do not modify the collection definition until you click Create or Modify to update the entire collection definition Enter Enter lets a new node to be added to the node list When you type in a new node name you can click Enter or press Return to add it to the list Modify Modify takes changes you make to a node definition and puts the changes into the list Modify is disabled until you have selected a node from the list by double clicking on the node Changes to the collection definition take effect when you click Modify at the bottom of the window Delete Delete removes any selected or highlighted nodes from the list The nodes are removed from the collection definition when you click Modify at the bottom of the window Chapter 8 Use the DECwindows Motif Interface 299 How You Customize Hot file queue To enable hot file collection in the CPD collection definition the toggle button to the left must be on This lets you specify the queue length in the text entry box This pertains only to the CPD collection definition because only the primary data collector collects hot file data Bl Hot file ques 0 333 Start dnte 7 OCT
473. ormance Manager Color ReGIS graphs The equivalence string must be a sequence of 4 plane definitions The following example demonstrates the setting for the default graph colors DEFINE PSPASHLS HOLOSO HOL50550 H160L42S100 HOL100S0 In this example plane O is black LO lightness zero and plane 3 is white L100 lightness 100 H is HUE 0 360 S is SATURATION 0 100 Appendix B Performance Manager Logical Names 407 PSPASPIE FONT POINT PSPASPIE FONT POINT Define this logical name to a number specifying a font point size for the PostScript pie chart item labels If the pie chart labels are so long that they extend beyond the sides of the paper use this logical name to cause the labels to fit on the page By default PSPASPIE FONT POINT is set to 10 Setting it to 7 should reduce the font sufficiently PSPASPS RGB 1 through PSPASPS RGB 6 These logical names allow the setting of the RGB color settings for the Performance Manager Color PostScript graphs You can also use them to specify grey shades for black and white printers Each logical name pertains to one of the 6 colors that may appear PSPA PS RGB 1 refers to the first color appearing at the bottom of the graph whose legend is located at the lower right of the display Specify a triplet of decimal values separated by spaces in the range of 0 1 where the first is for red then green and blue A lower value produces a darker shade For example the following settin
474. ort pull down the File menu and release on the Close menu item or click the Close button at the bottom of the window 278 Performance Manager Administrator Guide How You Display Analyzed Data Full Analysis Report Release on the Full Analysis menu item from the Display menu to open a Performance Manager Analysis Report window as shown in the following screen voum Start Time 26 AN 19937 09 00 End Time 26 JAN 1997 10 00 Conclusions 10070 There are many 1 0 operations or a high I O throughput te one or more disks The heavy usage of the diskis nay be the cause of an 1 0 bottleneck Suggested remedies include 1 Try to spread the I O demand over more disks workload to the noted 2 fry not to acd additional AwvV DISK OVER THRESHOLD CQ 1 00 PERCENT CPU TIME IN FILESYSTEM LT 30 00 Exec LT 35 00 OI r P OINO IO GATE OISK OVER THRESHO OI 5uAPP1N0 10 FAT DISK OVER THRE LT 015k 10 RAT D15k OVER THRESHOLD 2 00 TOP DIRIO PROCESS DIRIO GT OIRECT IO RATE Volume w Migh hst IO rate Itin itia Process u Highst Osk1O rate Time of I0 sec QLen sys mode C occurrence 34 18 1 16 2 11 3 00 scc 26 JAN 09 16 NO file of process CC ROB 1 3 67 1 22 3 3 CON CLIENT SCC 26 JAN 09 20 0 NO file of process SCCON D5 ROB 36 40 1 31 2 03 3 SCCON CLIENT SCC KO file of process SCCON 08 RO8 5 36 28 1 36 1 38 2 88 SCCON CLIENT a sss The menu bar contains File and View menus To
475. ort generation e 234 235 COMPILE command e 176 DISPLAY command e 178 DISPLAY INVESTIGATE displays e 391 EXPORT command e 182 GRAPH command e 191 PIE CHART command e 232 REPORT command e 234 Analysis and reporting facility e 15 Analysis report e 21 29 266 advice about performance e 18 contents e 235 effects of rule firing e 154 interpret e 22 order of conclusions e 21 rule evaluation of e 137 identifier e 21 selection of e 274 use auxiliary knowledge base e 135 173 window e 278 ANSI keyword e 191 Archive graphing from archived data e 230 Archiving schedule e 317 setting of e 317 AST e 84 Asterisk in histogram e 65 Auto augmentation e 319 Automatic scaling setting strip chart e 348 Auxiliary knowledge base e 236 building e 176 rules defined e 135 Bad list e 79 Balset count e 75 Bar graph setting high threshold e 347 low threshold e 347 maximum data value e 347 minimum data value e 347 patterns and colors e 350 peak hold units e 347 specifying fonts e 353 parts e 352 Baseline model e 15 Binary graph data loading of e 264 Bit map e 82 Block transfer e 85 Boolean operators e 153 Brief analysis report contents e 235 Buffered I O operation defined e 34 Buffrd I Os e 81 Build a knowledge base ADVISE PERFORMANCE command e 175 See also Knowledge base e 175 Build mode enabling of e 346 C Cluster data e 93 Disk e 101 rule description e 29 heading descriptions e 29 Server e 104 Clu
476. ory data for PA reports and graphs Process data and disk data can be filtered Performance Selection Filters Usernames to inchude ncn norrman roay i Imagenames to include 770700 Processmames to include IEEE Accoumtnames to inchede CF Uses to imchede eS Pids to include Dp Workloadnames to include CF Volumenames to include AA ox Cancel net Process data can be filtered by using any of the filter entries Usernames Imagenames Processnames Accountnames UICs PIDs or Workloadnames If a process record s identification information matches any of the identification specified that record is selected When using one or more of the process filters the following PA reports and graphs include only information on the selected processes m Process Statistics section of the Performance Evaluation Report m Process and Extended Process Metrics sections of the Tabular report m Hot file report only files used by the specified processes are selected m Top or Custom Username Imagename and Workloadname graphs m Top or Custom User Volume Image Volume and Workload Volume graphs m Top or Custom Hot file graphs only files used by the specified processes are selected 272 Performance Manager Administrator Guide How You Select Data for Analysis Likewise disk data can be filtered by using any of the filter entries Volumenames and Devicenames If a device record s identification information matches an
477. ou Select Data for Analysis see page 266 for information on requesting an analysis of performance 2 Ifthe analysis process is currently active the completion percentage is less than 100 percent Click the Cancel Operation button to stop work in progress and cancel data selection No reporting capabilities are allowed until a complete data selection is performed 3 When work is complete click the Dismiss button to clear the dialog box from your screen The Main Window s Data Selected for Processing Section is updated to reflect the new selection If data is missing from the selected time period missing data messages appear in the Work in Progress box This is to alert you that a subsequent examination of the selected analysis period may be incomplete or inaccurate The length of time required to read and analyze the selected data depends on the options selected such as hot file information and the duration of time 1 hour as opposed to one day You may want to remove the Main window and leave only the Work in Progress box on the screen Chapter 8 Use the DECwindows Motif Interface 263 How You Control the DECwindows Interface Read the Parameter File To read the parameter file 1 Release on the Read Parameter File menu item You are prompted with the following message b24 Unicenter PM cautions Important By clicking YES any changes to the parameter file that you have made are not saved and are lost
478. p first and what values to use for the Investigate screen scale factors The options are as follows m SCREEN NODE PROCESS DISK RULES INVESTIGATE SYSTEM INVESTIGATE MEMORY INVESTIGATE CPU INVESTIGATE IO INVESTIGATE LOAD RESOURCE MEMORY RESOURCE CPU RESOURCE DISK m PROCESS SCALING n 180 Performance Manager Administrator Guide ADVISE PERFORMANCE DISPLAY m WORKINGSET SCALING n m RATE PER SECOND SCALING n Chapter 6 Performance Manager Commands 181 ADVISE PERFORMANCE EXPORT ADVISE PERFORMANCE EXPORT Format Description Qualifiers The ADVISE PERFORMANCE EXPORT command allows Performance Manager data to be converted into a format you can use for further processing ADVISE PERFORMANCE EXPORT The EXPORT command converts Performance Manager data into a format that can be read and processed by an alternative analysis tool This command generates a file in CSV comma separated variable format Other tools can then import this data file directly for further processing BEGINNING time Specifies the starting date and time of the data to be exported If ENDING is not specified the default BEGINNING time is TODAY If ENDING is specified the default BEGINNING time is 00 00 of the date specified with ENDING BEGINNING is incompatible with the DATES qualifier CLASS item Specifies which optional classes of statistics are included in the export file You can
479. panel instruments e 342 Reporting period set e 269 Reports e 88 analysis report e 29 histograms e 64 70 mailing e 16 performance evaluation report e 33 62 how you interpret e 34 62 types of e 15 235 Residence time histogram e 235 image residence histogram e 64 performance evaluation report e 34 Response e 389 586 Performance Manager Administrator Guide time highest disk device e 191 highest files e 191 RESPONSE TIME graph type e 191 Rule cluster rule e 191 compiler e 176 conclusion definition e 156 condition accessibility of data cells e 156 definition e 152 display debugging information e 150 position within a rule construct e 148 request a list of e 21 string operators e 153 time data cell unavailable e 159 use string data cells e 159 constructs e 137 cpu rule e 191 element list of elements e 148 evidence accessibility of data cells e 156 example e 148 expression as a scan routine argument e 159 factory rules e 176 factory rules embedded in the product e 135 firing e 137 effects of rule occurrence e 154 identifier e 236 analysis report e 21 brief analysis report e 30 conventions e 149 definition and format e 149 position within a rule construct e 148 IO rule e 191 memory rule e 191 occurrence e 236 brief analysis report e 30 definition e 137 154 other rule e 191 pool rule e 191 resource rule e 191 user rules e 176 Rule modification e 18 Rules compiler e 15 invoke e 176 file auxiliary e 173 cont
480. pe The disk I O threshold such as TD T21 RAS81 is of the form TD Tn xxxx where n is the integer disk type as defined in STARLET and xxxx is the disk model name The data cell refers to the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP ARRIVG DECNET PACKET RATEMET F ARRLOCPK Average DECNET arriving local packet rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP AVERAGE IRPS INUSE Derived This contains a value representing the average number of IRPs in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY AVERAGE LOCKS INUSE Derived This contains a value representing the average number of locks in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY AVERAGE LRPS INUSE Derived This contains a value representing the average number of LRPsin use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY 414 Performance Manager Administrator Guide Performance Manager Data Cells AVERAGE RESOURCES INUSE Derived This contains a value representing the average number of resources in use on the local node for all of the intervals Data Type NUMERIC Domains SUMMARY AVERAGE SRPS INUSE Derived This contains a value representing the average number of SRPs in use on the local node for all
481. pe e 191 TOP USER VOLUME IO graph type e 191 TOP VA IMAGES graph type e 191 TOP VA USERS graph type e 191 TOP VA WORKLOADS graph type e 191 TOP WORKLOAD IMAGE ACTIVATIONS graph type e 191 TOP WRITE DISKS graph type e 191 TOP WRITE FILES graph type e 191 TOP WRITE VOLUMES graph type e 191 TOP WSSIZE IMAGES graph type e 191 TOP WSSIZE USERS graph type e 191 TOP WSSIZE WORKLOADS graph type e 191 Trans faults e 79 U Unlock parameters file e 259 schedule file e 259 User e 77 CPU panel e 327 faults panel e 332 memory field in system configuration report e 72 panel e 334 name e 309 V Video graphics e 373 Virtual Memory Requirements estimating e 535 pagelets e 535 Volume info panel e 335 VT100 e 373 VT240 e 373 W Warning message e 403 WCB e 81 Weekly schedule file modifying e 191 WIDTH keyword e 191 Window hits 81 interface for Performance Manager e 257 interface for Real time displays e 321 turns e 81 Working set e 298 Workload e 15 18 40 classification e 236 definitions base priority e 309 by sets of items e 309 creating e 309 defining e 308 deleting e 312 image e 309 image names e 309 included processes e 309 match both e 309 match either e 309 matching rquirement e 309 modifying e 308 312 PID e 309 process name e 309 transaction units e 309 UIC group e 309 user e 309 user names e 309 highest buffered IO e 191 cpu e 191 direct IO e 191 disk operations e 191 faulting e 191 har
482. pecifies that a pie chart representing data in units of percentages is to be filled out to be the MAXIMUM of 100 percent or is to represent only the TOTAL of the parts For example if you are producing a pie chart of CPU Utilization and the parts of the pie chart have the following values m Interactive 30 m Batch 10 m Network 5 m Overhead 1 m Interrupts 5 m Other 0 If you specify PERCENTAGE TOTAL the pie chart represents the sum of these parts a total of 51 percent utilization with the largest slice of the pie approximately 3 5ths being represented by Interactive If you specify PERCENTAGE MAXIMUM the pie chart contains a slice representing IDLE at 49 percent of the total pie with the remaining 51 percent representing their respective slices If PERCENTAGE is not used on the Pie command line then PERCENTAGE MAXIMUM is assumed This qualifier has no effect on graphs custom pie charts or pie charts of metrics other than CPU Utilization ADVISE PERFORMANCE REPORT Format The ADVISE PERFORMANCE REPORT command generates Analysis Reports Performance Evaluation Reports Tabular Reports and Histograms using daily or historical data ADVISE PERFORMANCE REPORT report_keyword 234 Performance Manager Administrator Guide Description ADVISE PERFORMANCE REPORT Use the ADVISE PERFORMANCE REPORT command to produce Performance Manager reports The Performance Manager can generate reports using either daily
483. ph for System Metrics Chapter 4 Generate Historical Graphs 127 Generate Custom Graphs Processing Options ANALYSIS REPORT NO PERFORMANCE EVALUATION REPORT NO GRAPHS YES User Names YES Image Names YES Hot File Names NO User Image Volume IO NO IO Devices amp workloads YES By Node details YES Reading data for node YQUEM Reading data for node GALLO PSPA GRAPH TYPE PROMPT FORMAT POSTSCRIPT CHARACTERISTIC PATTERN OUT CHACUSTOM PROMT PS Please select either 1 a predefined graph or 2 a custom graph Choice 1 2 0 Composite 1 YQUEM 2 GALLO Please select a node number 0 2 For the CUSTOM Graph select one of the following System Metrics Process metrics selected by user name Process metrics selected by image name Process metrics selected by workload name Disk device metrics selected by device name Disk device metrics selected by volume name Processor mode metrics by physical processor ID HSC metrics by HSC node name SCS metrics by SCS Cluster node name RULE metrics by rule id HSC Channel metrics by channel name 12 File metrics by file name 13 I0s by Username and Volumename 14 IOs by Imagename and Volumename DO O NO UI i UN FH H HO Enter Choice 1 14 1 1 Select Up to 6 System Metrics lt CR gt for list 2 GRAPH TYPE CUSTOM System metrics Select up to 6 of the following system metrics to be displayed on a custom graph Sampled CPU mode metrics for all CPUs in an SMP
484. phs Node Control and Toggle Buttons By default the Performance Manager analyzes data for all nodes in your schedule file Click a node s toggle button to include or exclude the node from the processing Clear and Set buttons are available for adjusting all toggles Control Buttons The OK button applies your selection removes the data selection box and activates a Performance Manager Work in Progress dialog box which includes a real time display of the progress of the analysis procedure See the Controlling the DECwindows Motif Interface section for details of the Work in Progress box The Cancel button closes the Data Selection dialog box and resets all widgets The Reset button sets all widgets back to their settings of either the startup defaults or the last approved selection Select the Last Hour Release on the Last Hour option from the Select menu to apply the current data selection settings for the last hour of the day To review what the current data selection settings are release on the Specific data menu item and view the Data Selection dialog box Use Custom Default Settings You can override the default data selection settings by providing a selections resource file PSPASSELECT DAT in the DECW USER DEFAULTS directory area typically SYS LOGIN A sample file is provided in PSPASEXAMPLES The PSPASSELECT DAT file will be read once when you make the first data selection This file can be edited with a text edit
485. phs from the DCL Level Generating graphs at the DCL level offers the most efficient method of selecting data Only those metrics contributing to the selected graphs are saved and bucketed when the performance data is read The CPU demand is mostly for reading and decoding the data files By contrast the data selection for the DECwindows and command mode interfaces is somewhat more costly because all the performance metrics are usually saved and bucketed in anticipation of a subsequent user request to view them See Appendix D for information on estimating virtual memory needs and selecting data To generate a graph or pie chart m Use the following commands ADVISE PERFORMANCE GRAPH ADVISE PERFORMANCE PIE CHART With either command you can control the data selection by specifying a time period and list of node names The TYPE qualifier specifies graphs or performance metrics or both The FORMAT qualifier controls the format of the output data The OUTPUT qualifier directs the output data to the desired destination You can generate all of the predefined graphs in a single command by using the TYPE ALL GRAPHS qualifier However if you do not specify a graph the CPU Utilization graph is the default Generate Graphs in Command Mode If you are investigating performance data interactively generating graphs from command mode offers several distinct advantages Once the data has been loaded into memory you can view graphs and reports
486. play This shows the working set size and page fault rate for the top faulting processes Adjust the scaling factors if necessary Look for processes that are faulting heavily but have small working sets If your system has ample memory increase the working set quota WSQUOTA and the working set extent WSEXTENT for these processes If memory is short on your system increase WSQUOTA and WSEXTENT for these processes at the expense of processes that are not faulting but have large working sets Chapter 11 Use the Character Cell Real time Display 397 Evaluate Performance Using the Investigate Displays Inappropriate automatic working set adjustment AWSA parameters Look at the Process bars at the right on the Memory display Look for top faulting processes with fluctuating working set sizes If the working set size for such a process increases and decreases accompanied by page faulting then the AWSA parameters might be out of adjustment System parameters that affect automatic working set adjustment are PFRATH PFRATL WSINC WSDEC AWSTIME AWSMIN GROWLIM BORROWLIM and QUANTUM Automatic decrementing can be turned off by setting PFRATL 0 this is normally recommended Do not change any of the other parameters without a thorough understanding of the AWSA mechanism The automatic memory reclamation mechanism of OpenVMS should be enabled This is controlled with the SYSGEN parameter MMG CTLFLAGS Too many image activations Look at
487. previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous TD I82 RFH35 TD I83 RF31F TD I84 RZ72 TD I85 RZ73 TD I86 RZ35 TD I87 RZ24L TD I88 RZ25L TD I89 RZ55L TD I90 RZ56L TD I91 RZ57L 2600000 1400000 1400000 1900000 2600000 1900000 2000000 1100000 1500000 1800000 Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Name TD 192 RA73 TD 193 RZ26 TD I94 RZ36 TD I95 RZ74 TD I96 ESE52 TD I97 ESE56 TD I98 ESE58 TD I99 RZ27 Default Value 2100000 2600000 3700000 3100000 2000000 2000000 2000000 3700000 Components of Rules Description Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Same as previous Processor Specific Thresholds The processor specific thresholds can be modified by using threshold scaling factors indicated in the following table Threshold Name TD SINGLE CPU VUP n TD SOFT FAULT SCALING n TD HARD FAULT SCALING n TD IMGACT SCALING n TD COM SCALING n Affected Data Cell Name CPU VUP RATING SOFT FAULT SCALING HARD FAULT SCALING IMGACT SCALING COM SCALING The n is an integer that specifies an individual processor and must be
488. priority 0 to 31 This is the priority of the process at the time the example is taken State The process scheduling state This is the state of the process at the time the example is taken In final tabular statistics the state of the last interval is reported Image Count The number of images activated by the process during the interval 90 Performance Manager Administrator Guide Tabular Report Sections CPUtime min The CPU time in minutes accrued by the process during the last reporting interval Direct I O The number of direct I O operations issued by the process during the last reporting interval as a rate per second Buffrd I O The number of buffered I O operations issued by the process during the last reporting interval as a rate per second Page Fits The number of page faults incurred by the process during the reporting interval as a rate per second FIt I O The number of page fault I Os incurred by the process during the reporting interval as a rate per second Working Set MIN AVE MAX The minimum average and maximum working set size for the process during the reporting interval Mo The process mode IN Interactive BA batch NE network DE detached Image Name A line containing the image name follows each process metrics line if the image name was collected and reported This is the name of the image at the time the example is taken Chapter 3 Evaluate Performance in Detail 91
489. process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO Appendix C Performance Manager Data Cells 483 Performance Manager Data Cells PROCESS DISK THRUPUT TALLY Derived This contains the sum of the values representing the disk throughput for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS IMAGE ACTIVATIONPRO B IMGACT This contains a Boolean value where for CPD data a 1 means the process s activated the image during the current interval and for History data a 1 means that one or more activations took place for data summarized into this process subrecord This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS IMAGE ACTS TALLY Derived This contains the sum of the values representing the image activation rate for all of the current process subrecords which were selected by the most recent PROCESS SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS_IMAGE_ACT_RATEPRO_F_IMGACTS This contains a value representing the image activation rate of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS_IMAGE_LOGINPRO_B_LOGIN This contains
490. pter Generate Historical Graphs see page 107 Graph Window File Edit Display Display Top Display Custom CPU UTILIZAT Monochrome Specifies that shades of a single color shall be used to display the graph or pie chart Chapter 8 Use the DECwindows Motif Interface 289 How You Display Analyzed Data Line Mode Specifies that the values for each category on the graph are to be displayed in a line mode Click the button to turn line mode off or on File Edit Display Display Top Display Custom Editing Panel Release on the Editing Panel menu item and the Performance Manager interface displays a Graph Panel dialog box Graph Selection Y Axis Maximum 10 a Operation Rate Top Disk Devices o Vome col By ls Threshold Line Max 72 Min 7 2809 Ave 33 vae 33 290 Performance Manager Administrator Guide How You Display Analyzed Data The Graph Panel dialog box contains five sections that allow you to enter any of the following Graph selection The first option menu lets you specify a major class of graphs to be viewed In response Performance Manager replaces the entries in the second option menu with those that are appropriate for the class you specified in the first option menu The graph window is updated with the first graph in the list indicated by the major category An exception to this is switching between the Top User a
491. pts per second that were made to find extents in the extent cache Bit Map The number of attempts per second that were made to find entries in the bit map cache File Cache Effectiveness For each item for example Dir FCB the effectiveness is computed as the ratio of item hits item hits item misses expressed as a percentage Dir FCB The effectiveness of the directory cache for finding directory file control blocks Dir Data The effectiveness of the directory cache for finding directory data Quota The effectiveness of the quota cache File Id The effectiveness of the file ID cache File Hdr The effectiveness of the file header cache Extent The effectiveness of the extend cache Bit Map The effectiveness of the bit map cache Chapter 3 Evaluate Performance in Detail 83 Tabular Report Sections Lock Rates This section reports the lock manager metrics collected from OpenVMS performance statistics This report contains three columns Local In and Out Each of these columns report rates for a variety of lock manipulation requests such as New ENQ Converted ENQ and DEQ Rates for locking information are Local Lock manipulation requests made at the local node for the benefit of that node In Lock manipulation requests coming to the local node from other nodes in a cluster Out Lock manipulation requests being sent from the local node to other nodes in the cluster Lock manipulation reque
492. r You can set the bar graph with the following attributes Background Normal Range Low Threshold High Threshold Peak Hold 350 Performance Manager Administrator Guide How You Edit the Panel Instruments To set the Patterns and Colors for each of these attributes 1 Click the appropriate button Performance Manager displays the Pattern Editor dialog as shown in the following screen Pattern Selection Pattern Colors reresromnd mm Background 2 Click a pattern in the dialog and the selected pattern is displayed in the pattern viewer To change the graph s pattern color click the Foreground or Background button Performance Manager displays a color mixing dialog 3 Click Help for information on how to use the color mixing dialog Strip Chart Strip Chart Patterns and Colors Chart Background Normal Range Low Threshold High Threshold Chapter 10 Customize the DECwindows Motif Real time Display 351 How You Edit the Panel Instruments The Strip Chart Patterns and Colors dialog lets you set the pattern and color of the chart with the following attributes m Background m Normal Range Low Threshold High Threshold To set the Patterns and Colors for each of these attributes 1 Click the appropriate button Performance Manager displays the Pattern Editor dialog 2 Click a pattern in the dialog and the selected pattern is displayed in the pattern viewer 3 To change the chart
493. r Format In the By Cluster format for disk statistics the leftmost column of the tabular report contains the name of each disk that is accessible cluster wide Disk names are usually prefixed by a node name or allocation class Each disk name begins a line of disk statistics Each statistic shows the activity for the specific disk due to all nodes in the cluster When determining cluster wide statistics for a disk only statistics from nodes with direct access are considered This is because I O for nodes with remote access is processed by nodes directly accessing the disk and is therefore already included in the node with direct access statistics For example the software would calculate cluster rates for disks DU1 and DB2 in the following figure as follows m For disk DU1 with the HSC all I Os come to this disk through host direct access nodes A and B Therefore adding the I O statistics from each of these nodes provides the total cluster I O for disk DU1 The I O statistics of node C are already included in the I O statistics for nodes A and B and are therefore ignored Chapter 3 Evaluate Performance in Detail 101 Tabular Report Sections m For disk DB2 node C is the host direct access node All I Os come to this disk through node C therefore the I O statistic of node C provides the total cluster I O rate for disk DB2 As above the I O statistics of nodes A and B are already included in the I O statistics for node C and are th
494. ram The following example shows a terminal I O per second histogram Histograms DEMAND VAX 8700 Page 8 PA Vx x Tuesday 06 SEP 2005 00 00 to 23 59 TERMINAL I 0 PER SECOND per sec 2 L k L LL LE CEL L LL L LLLLL RLLL LLLEL LLE L LLLLLLLLLRLLLLLLLLLLLLLL LL LLLLLLLLLLLRLLLLLRLLLLLLLLLLLL LLLLLLLLRLLRLLLLLLLRLLLLLRRLRLLLLRLLLLLLLLLLLLLL LL 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 3 Each Column represents approximately 20 minutes starting from 6 SEP 00 00 00 to 6 SEP 23 59 00 An N indicates NO DATA The following statements are keyed to the columns in the previous example 1 Explanatory list of units in the histogram columns 2 The number of terminal I Os per second 3 Number of hours spanned by the reporting period This histogram reflects a reporting period of 24 hours 68 Performance Manager Administrator Guide Histograms System Uptime Chart Histogram The following example shows a system uptime chart histogram Histograms CLUSTER Page 13 PA Vx x Tuesday 06 SEP 2005 00 00 to 23 59 The following chart presents the status of each node in the cluster over the report time period Legend 1 Node up with data Node up no data wanted Node up no data found Node down unknown and no data 2 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 3 Each Column represents approximately 20 minutes starting from 6 SEP 00 00 00 to 6 SEP 23 59 00
495. ransferred to or from the disk 5 Average size of the I O operations to the disk in pages The average size for each disk is followed by a breakdown of this value from contributing nodes when applicable 6 Name of the node that uses the various disk volumes For rows in which no source node name appears the data refers to the cluster wide activity on the disk volume 7 Name of the node that services I O requests An asterisk denotes that more than 1 server existed for this disk 8 Percentage of time that I O requests are outstanding to the volume for each node utilizing the disk volume 9 Percentage of total I O activity devoted to read operations This value for each disk is followed by the percentage of read operations to the disk from contributing nodes when applicable 10 Percentage of total I O activity that were split I O operations This value for each disk is followed by the percentage of read operations to the disk from contributing nodes when applicable 11 Disk volume type In this example all but one of the disks are RA81s 12 Count of the Performance Manager records containing data for a disk volume during the interval Interpret cluster wide Tape Statistics The tape statistics follow the cluster wide disk statistics To display only the tape statistics section of the Performance Evaluation Report use the following qualifier INCLUDE TAPE STATISTICS To disable the tape statistics display from the
496. ration data has been gathered Data collection started The date time data collections were started Data collection ended The date time of the last record in the log file Example interval The interval at which data is collected also referred to as the collection interval The collection interval is expressed in seconds and minutes Data is collected during each interval and written to the log file at the specified intervals Report generated The time when the report was generated Processor type is The VAX processor type for example 6000 440 72 Performance Manager Administrator Guide Tabular Report Sections Running OpenVMS version The current version of OpenVMS Total memory The total physical memory used by the OpenVMS operating system in pages and in megabytes This is the smaller of the actual physical memory on the system and the system parameter PHYSICALPAGES For Integrity and Alpha systems memory figures are presented as pagelets 512 bytes Nonpaged memory The difference between the amount of total memory and paged memory total memory paged memory expressed in pages and as a percentage of total memory It includes the PFN database nonpaged executive code and data nonpaged pool and the system header If the system parameters POOLPAGING or SYSPAGING are set to zero then the paged pool or paged system pages which are normally paged in the system working set are instead allocated as nonpaged memory
497. rator Guide REPORT Description The Performance Manager software can produce a multitude of predefined or custom pie charts in the following formats PostScript and DECwindows formats tabular and CSV In command mode the OUTPUT qualifier must be used to direct the output to a PostScript file The PERCENTAGE qualifier is specific to the PIE CHART command For other applicable PIE CHART qualifiers see the GRAPH command see page 245 Qualifiers PERCENTAGE TOTAL MAXIMUM Specifies that a pie chart representing data in units of percentages is to be filled out to be the MAXIMUM of 100 percent or is to represent only the TOTAL of the parts For example if you are producing a pie chart of CPU Utilization and the parts of the pie chart have the following values m Interactive 30 m Batch 10 m Network 5 m Overhead 1 m Interrupts 5 m Other 0 If you specify PERCENTAGE TOTAL the pie chart represents the sum of these parts a total of 51 percent utilization with the largest slice of the pie approximately 3 5ths being represented by Interactive If you specify PERCENTAGE MAXIMUM the pie chart contains a slice representing IDLE at 49 percent of the total pie with the remaining 51 percent representing their respective slices If PERCENTAGE is not used on the Pie command line then PERCENTAGE MAXIMUM is assumed This qualifier has no effect on graphs custom pie charts or pie charts of metrics other than CPU
498. rcent of CPU Time Other Network Batch Interactive Detached Intstk MPsynch 19 FEB 2006 13 00 0417 8618 34 8204 34 7315 29 2958 7 1025 19 FEB 2006 13 05 7843 6284 25 25 5536 13 9655 7340 19 FEB 2006 13 10 8503 6480 29 29 0197 26 4432 9179 F4 8 19 FEB 2006 13 15 5400 1330 37 37 2375 35 1232 9 5303 19 FEB 2006 13 20 6206 3342 16 16 1200 7037 9 1339 19 FEB 2006 16 19 FEB 2006 16 19 FEB 2006 16 19 FEB 2006 16 19 FEB 2006 16 19 FEB 2006 16 Minimum Values Maximum Values Average Values BUM Ui Ui O Nn The following example CSV Formatted Output illustrates CSV formatted output It was generated with the following command ADVISE PERFORMANCE GRAPH BEGINNING 22 MAR 2006 12 00 ENDING 22 MAR 2006 14 00 _ NODE BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY FORMAT CSV X POINTS 8 _ TYPE CUSTOM SYSTEM METRIC CPU TOTAL COMPOSITE PSPA CUSTOM GRAPH Nodes BYOB BUGDEV ORIPAS BATCH MIGHTB CINAMN CHATTY Date 22 MAR 2006 12 00 14 00 Metric Values are Stacked Added to the left Units CPU Total Busy Time 2 Other Nodes total BUGDEV BYOB CINAMN ORIPAS BATCH 22 MAR 2006 12 00 4 5250 3 8487 2 7485 1 5865 0 9080 0 6247 22 MAR 2006 12 15 4 3510 3 6925 2 9084 1 2251 0 4400 0 1722 22 MAR 2006 12 30 3 5105 2 9182 2 7072 1 5696 0 9082 0 6624 22 MAR 2006 12 45 4 6329 4 0381 3 6512 2 6581 1 2887 1 0471 22 MAR 2006 13 00 9 6964 9 1100 2 8984 1 8634 0 8863
499. re to use for the graph If you omit this qualifier daily data is obtained from the Collection Definition called CPD To view the Collection Definitions that you have available use the DCL command ADVISE COLLECT SHOW ALL If you want to use history data instead of daily data use the HISTORY qualifier instead of the COLLECTION DEFINITION qualifier COLLECTION DEFINITION is incompatible with the HISTORY qualifier CLASSIFY BY USERGROUP family name Specifies the workload family whose workload definitions are to be used for summarizing process activity This affects the TOP WORKLOAD graph types as well as custom graphs with WORKLOAD metrics by providing the desired metrics on an individual workload basis The default is other which averages all process activity together The family type of USERGROUP is required No restrictions are made on the family name 192 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH COMPOSITE Combines data from all nodes into a single graph Data from each node is either added or averaged The following command produces a graph of the total number of processes in the cluster ADVISE PERFORMANCE GRAPH COMPOSITE TYPE PROCESSES When the Performance Manager combines I O data from more than one node it is possible to double count I O operations to a disk device if it is served Therefore when you specify COMPOSITE the Performance Manager does not count all MSCP served I O for
500. read I O operations per second to the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE SIZEMET F CACHE SIZE This contains the current size in pages of the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE USEDMET F CACHE USED This contains the number of used pages in the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 420 Performance Manager Administrator Guide Performance Manager Data Cells CACHE WBYPASSMET F CACHE WBYPASS This contains the number of write I O operations per second bypassing the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE WRITEIOMET F CACHE WRIO This contains the number of write I O operations per second to the I O Cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CHANNEL OVER THRESH PORT Derived This contains the adapter nexus number that is experiencing excessive throughput on the local node for the current interval or a zero It is zero if the Boolean data cell EXCESS THRUPUT ON ANY CHANNEL is zero Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CHANNEL_OVER_THRESH_THRUPUT Derived This contains a value representing the
501. received per second on the local node from the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG SEQUENCED_MESSAGES_RECEIVEDSCS_F_MSGRCVD SEQUENCED_MESSAGES_RECEIVED This contains the value representing the number of messages received per second on the local node from the remote Data Type NUMERIC 498 Performance Manager Administrator Guide Performance Manager Data Cells Domains CFG SEQUENCED MESSAGES SENTSCS F MSGSENT This contains the value representing the number of messages sent per second from the local node to the remote node for the current configuration record and interval Data Type NUMERIC Domains CFG SEQUENCED MESSAGES SENT TALLY Derived This contains the sum of the values representing the number of messages sent per second from the local node to the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG SMALLEST_BLK_IN_NONPAGED_POOLMET_F_NP_MIN_BLOCK This contains the number of bytes in the smallest block in non paged pool for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP SMALLEST_BLK_IN_PAGED_POOLMET_F_PG_MIN_BLOCK This cont
502. rem nodes deadlock msg sec sent to remote nodes Position Length 199 7 208 7 217 7 226 7 44 0 eng wait 0 0 eng notqd 0 0 dick search 0 0 dick find 10308 0 tot locks 4835 0 tot resources of times lock unavailable and process waited of times lock unavailable and process did not wait of times a deadlock search initiated by Icl system of times a deadlock condition was found by Icl system total of Icks outstanding total of resources that can be locked 235 7 244 7 253 7 262 7 271 7 280 7 552 Performance Manager Administrator Guide Device Statistics Data Record Device Statistics Data Record The following example commands show how to export and display device statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP DEV CLASS NODEFAULT DEVICE _ BEGINNING 26 JAN 1997 14 00 00 00 ENDING 26 JAN 1997 14 02 00 00 TYPE EXP DEV ULTRA VERS Vx x REPORT LOG FILE Vx x CPD 26 JAN 1997 14 00 00 00 26 JAN 1997 14 02 00 00 ULTRA DEVI 1 DADO 0 0 Example Data item Description Position Length 1 DADO device name device name 92 20 0 0 device rate I O 115 7 operations second to device Disk Statistics Data Record The following example commands show how to export and display disk statistics ADVISE PERFORMANCE EXPORT NODE ULTRA OUTPUT EXP DISK CLASS NODEFAULT DISK _ BEGINNING 26 JA
503. report use the following qualifier INCLUDE NOTAPE STATISTICS 58 Performance Manager Administrator Guide Performance Evaluation Report The following example shows the tape statistics section of the Performance Evaluation Report Performance CLUSTER Page 1 Evaluation PA Vx x Wednesday 14 JAN 2006 09 00 to 19 20 The following table gives the summary of all tape activity as seen by the indicated node An 4 for service node indicates that more than one was detected Tape Avg I O Avg Avg I0sz Source Service IO IO of Vo 1 2 3 4 2 6 7 8 9 10 11 12 8JULO2 2 MUA43 3 76 20 14 119 1 63 3 LATOUR JULIO 15 46 0 0 TA90E ABC 2 MUA43 99 21 61 30 270 5 5 5 LATOUR JULIO 39 62 0 0 TA90E The following statements are keyed to the columns in the previous report 1 Volume name of the tape to which one or more nodes in the cluster directs activity 2 Average number of I O operations per second to the tape volume For each tape the total I O per second is reported followed by a breakdown of this activity from contributing nodes when applicable 3 Average number of I O requests waiting for service to the tape The average queue size for each tape is followed by a breakdown of this value from contributing nodes when applicable 4 Average number of kilobytes per second transferred to or from the tape 5 Average size of the I O operations to the tape in pages The average size for each tape is followed by a breakdo
504. requested time period and nodes against the rules in the knowledge base The Analysis Report consists of conclusions for each node and includes cluster wide conclusions for a cluster system Each conclusion is caused by a rule firing When a rule fires the Performance Manager reports the problem condition and makes a recommendation for improving performance All rules are identified with a unique five character alphanumeric code such as M0010 A rule identifier appears with each conclusion In addition to the conclusions you can request that the Performance Manager list the rule conditions that satisfied the rule firing The Performance Manager also provides supporting evidence By default the Performance Manager provides conditions and evidence when you specify an output file or when you run the analysis process in batch mode To suppress the conditions and evidence in the report use the NOEXPLAIN qualifier The conclusions in the Analysis Report for a cluster system are listed in the following order by node m Each node s local analysis and conclusions which may include the following conclusions Memory related m I O related CPU related Miscellaneous Auxiliary Chapter 2 Analyze Performance 21 Analysis Reports m Cluster wide analysis and conclusions which may include the following conclusions HSC limitation Disk related Lock related Auxiliary If you request a Performance Evaluation Report alo
505. ric class is VOLUME METRICS then the selection strings are interpreted as volume names 204 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH Ifthe metric class is CPU METRICS then the selection strings are interpreted as physical processor IDs which are in the form NODENAME INTEGER such as NODE1 3 To display a graph which shows active CPUs in an OpenVMS multiprocessing system enter a command similar to the following ADVISE PERFORMANCE GRAPH END 0 10 NODE YQUEM _ TYPE TOP BUSY PROCESSOR Specifying a physical CPU ID allows you to isolate and analyze one CPU of a selected node in an SMP configuration Ifthe metric class is HSC METRICS then the selection strings are interpreted as HSC node names Ifthe metric class is SCS METRICS then the selection strings are interpreted as cluster node names Ifthe metric class is FILE METRICS then the selection strings are interpreted as file names Ifthe metric class is RULE METRICS then the selection strings are interpreted as rule IDs Note Rule Metrics are available only from history files Ifthe metric class is DISK USER METRICS then the selection strings are interpreted as username volumename Ifthe metric class is DISK IMAGE METRICS then the selection strings are interpreted as imagename volumename Ifthe metric class is CHANNEL METRICS then the selection strings are interpreted as a channel spec which is the HSC nodename an underscor
506. rn performance analysis When you request an Analysis Report you may receive advice about performance problems for some of the following reasons m Inefficient setting of SYSGEN or UAF parameters m Excessive demand m Excessively long device queues m Insufficient system resources m Inefficient application design m Insufficient hardware for the workload Before you follow any Performance Manager recommendations ask the following questions m Is the problem caused by a temporary condition Performance Manager does provide summary analysis for extended time periods such as weeks or months m How frequently does the problem occur m Is there a difference in the workload before and during the time the problem occurs Occasionally the problem may be caused by the inherent characteristics of an application or workload If Performance Manager recommends that you need additional hardware keep track of this recommendation over a period of time If system performance degrades this recommendation occurs more frequently You must decide whether the problem is serious enough to warrant additional hardware 18 Performance Manager Administrator Guide What to Expect from Performance Manager Never rely solely on Performance Manager recommendations As you become familiar with the workload on your system you develop your own ideas on how to recognize and alleviate performance problems Use the Performance Manager as a tool to help you
507. rrent interval Data Type STRING Domains PRO 526 Performance Manager Administrator Guide Performance Manager Data Cells VBS INTSTKMET F VBSSCPUTICK Average percentage of CPU time on the Interrupt Stack spent on behalf of VBS Virtual Balance Set transitions only for all processors in the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP VMS543 OR LATER Derived This contains a one if the version of OpenVMS is V5 4 3 or later zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP VMS60 OR LATER Derived This contains a one if the version of OpenVMS is V6 0 or later zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP VMS732 OR LATER Derived This contains a one if the version is OpenVMS 7 3 2 or later zero if not for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP VMS82 OR LATER Derived This contains a one if the version is OpenVMS 8 2 or later zero if not for the loca
508. rule condition you must compare the string data cell to another string data cell or to a string literal with one of the string operators EQS or NES The time data cell represents the time of the current Performance Manager interval record You cannot use this data cell in a rule condition however it is valid as an evidence item Scan Routine Data Cell You can use scan routines to scan sub records in the target domain The scan routine data cell name starts with the target domain name The scan routine data cell counts sub record occurrences and tallies data into tally data cells for use in a rule expression The scan routine data cell requires a rule expression enclosed with parentheses following the scan routine name This rule expression can contain data cells These cells referenced in the rule expression must be in the target domain scanned by the scan routine For example you can use the scan routine data cell PROCESS SCAN to return tally information If a rule is in the LOCAL domain you can use the following rule expression to test whether the image XYZZY EXE is being used PROCESS SCAN IMAGE NAME EQS XYZZY GT 0 The PROCESS domain is the target domain for PROCESS SCAN The value that a scan routine returns is an integer indicating the number of times the specified expression tested true evaluated to 1 0 In this case the value returned by PROCESS SCAN would be the count of process sub records with an image nam
509. ruput Hard Fault Rate Soft Fault Rate Buf I O Rate CPU Utilization list Total Mem Pages WSdefault WSquota WSext Private Global Metric Disk I O for this image Disk throughput Hard page fault rate Soft page fault rate Buffered I O rate CPU Utilization for image by user Total of all Working Sets for processes running this image Working set default setting for this process Working set quota setting for this process Working set extent setting for this process Private memory size for this process Global memory size for this process Units Operations per second Kilobytes per second Faults per second Faults per second Operations per second Percentage Pages Pages Pages Pages Pages Pages Next Panel Hard Fault Rate Panel Descriptions The Fault Rate graph on the System Overview panel lets you access the following panels m Page Faults m User Faults m Process Faults m Fault Types m Memory Memory Allocation 330 Performance Manager Administrator Guide Page Faults Panel Faults Rates Panel m User Memory m Page File Allocation Default Panel Descriptions This series of panels lets you investigate the causes of hard page faulting on the monitored system You can investigate the top faulting and top memory users and view what processes each user is executing The following tables describe the information presented wit
510. s DISK QUEUE AT SERVER Derived This contains a value representing the OpenVMS node server queue length for the local disk during the current interval For an HSC based disk this cell This contains the highest queue for all nodes on the CI Data Type NUMERIC Domains CLU DISK QUEUE LENGTHDEV F QLEN This contains a value representing the average number of outstanding I O requests for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK QUEUE LENGTH TALLY Derived This contains the sum of the values representing the average queue length for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_QUOTA_CACHE_SIZEDEV_F_QUOSIZE This contains a value representing the number of entries in the quota cache for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_READ_IO_RATEDEV_F_RDCNT This contains a value representing the average number of read I O requests per second from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK Appendix C Performance Manager Data Cells 441 Performance Manager Data Cells DISK READ IO RATE TALLY Derived This contains the sum of the values representing the average read I O rate for the current disk subrecords which were selected b
511. s Displays options for summarizing process statistics and for specifying the reporting intervals as shown in the following screen Process Statistics Summarization Wy Afcount Name By Imagename W y Mode L By PID O By Process Name Ii By UIC Group W y Username W y workload Tabular Report Options O Process Statistics Chapter 8 Use the DECwindows Motif Interface 275 How You Select Data for Analysis Tabular Interval Directs the Performance Manager to prepare interval reports summarizing data into specified intervals This reporting option is unavailable unless the Final Tabular report is selected Once selected the reporting interval becomes available Performance Manager data collection for daily performance data is recorded every two minutes This is the default reporting interval This value can be adjusted by varying the supplied value of 2 and the units value of minutes When the analysis process is initiated it may override your selections for the reporting interval if it is not a multiple of the recording interval Graphs Directs the Performance Manager to analyze the data to generate all graphs and pie charts The options are as follows m By Imagenames Enables graphing of process statistics by imagename m By Usernames Enables graphing of process statistics by user name m By Hot Filenames Enables graphing of file statistics by file name m By Disks channels CPUs HSCs workloads SCS nodes and rul
512. s COM CFG CPU DSK FIL LOC PRO TAP FILE ID CACHE AR Derived Attempt rate per second to the file ID cache for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP FILE_ID_CACHE_HR Derived Hit ratio to the file ID cache for the local node for the current interval record Calculated by dividing the number of file ID cache hits by the number of file ID cache attempts hits misses times 100 Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP FILE_MSCP_IO_RATEFIL_F_MSCPOP This contains a value indicating the number of MSCP served I O operations per second issued to the file indicated by FILE_DEVICE FILE_DIRECTORY and FILE_NAME for the current interval Data Type NUMERIC Domains FIL 448 Performance Manager Administrator Guide Performance Manager Data Cells FILE NAMEFIL A FILE This contains a string indicating the name of the file for which the current hot file subrecord pertains in domain FILE Data Type STRING Domains FIL FILE OPEN RATEMET F OPENS Average file open rate per second for the local node for the current interval record Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP FILE OPERATION RATEFIL F OPCNT This contains a value indicating the number of I O operations per second issued to the file indicated by FILE DEVICE FILE DIRECTORY and FILE NAME for the current interval Data Type NUMERIC Domains FIL FI
513. s LOC CACHE MISSES 3364MET F CACHE MISS 3364 The number of read operations with a block size from 33 to 65 that bypassed the XFC for the current interval for the local node Data Type NUMERIC Domains LOC CACHE MISSES 65127MET F CACHE MISS 65127 The number of read operations with a block size from 33 to 65 that bypassed the XFC for the current interval for the local node Data Type NUMERIC Domains LOC CACHE MISSES 128255MET F CACHE MISS 128255 The number of read operations with a block size from 128 to 255 that bypassed the XFC for the current interval for the local node Data Type NUMERIC Domains LOC Appendix C Performance Manager Data Cells 419 Performance Manager Data Cells CACHE MISSES GT255MET F CACHE MISS GT255 The number of read operations with a block size greater than 255 that bypassed the XFC for the current interval for the local node Data Type NUMERIC Domains LOC CACHE RBYPASSMET F CACHE RBYPASS This contains the number of read I O operations per second bypassing the I O cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE READHITSMET F CACHE READHITS This contains the number of read I O operations per second to the I O Cache that were satisfied by the cache for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CACHE READIOMET F CACHE RDIO This contains the number of
514. s default by specifying values for individual nodes Default minimum space Enter the minimum number of blocks of free disk space needed on each database disk for each node in the definition Data collection suspends recording if there are fewer blocks of free disk space available You can override this default by specifying values for individual nodes Default database path The default database path for all nodes in the collection definition You can override this default by specifying values for individual nodes 298 Performance Manager Administrator Guide How You Customize Node Definition Section This displays the nodes in a collection definition Default values are displayed Although the changes that you make appear as they are made they are not applied to the collection definition until you click the Create or Modify buttons at the bottom of the window Node Working Set Minimum Space Database Path Enter YOUEN zeae 1590 PEOC SOATABASE LaTouR zeae 2500 PSDC IOATADASE rus vs tem 2490 PEOC TOATABASIT aa isa memet 1 reas 1594 gpsuc SpA TAnas t lal dtn w To add a node enter the name and make any modifications you want to the default values shown to the right Then press Return or click Enter To modify a node entry double click it The definition appears under the list Make the changes you wish and then press Return or click Modify to the left of the box To delete one or m
515. s are the source files containing user defined site specific rules and thresholds that collectively comprises the auxiliary knowledge base Baseline load A Baseline load is the Performance Manager s measurements of your existing system or cluster System This data is stored in a model input file MDL Glossary 565 Baseline model A baseline model is a model generated from historic or daily Performance Manager data The model output provides a workload characterization report An unmodified MDL file the baseline load represents Performance Manager measurements of your existing system The model generated from this file is known as the baseline model Buffered I O operation A buffered I O operation occurs each time an intermediate system buffer is used in place of the process context buffer Business workload A business workload is the work the business expects its computer to perform Computer system The computer system is all of the computer hardware on which business work is performed Conclusions Conclusions are Text displayed in an analysis report CPU branch explicit In a modeling context the CPU branch explicit is the probable distribution of load across CPUs for a workload or transaction class Probabilities must sum to one This data is included in the model input file Custom graph A custom graph is a graph type which allows you to specify which Performance Manager data items to graph Daily data files Daily dat
516. s contains a value representing the number of entries in the bitmap cache for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK Appendix C Performance Manager Data Cells 437 Performance Manager Data Cells DISK MAX BLOCKSDEV F MAXBLOCK This contains a value representing the maximum number of blocks on the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK MOST FULL X Derived This contains an index pointing to the disk subrecord experiencing excessively limited free space on a disk on the local node for the current interval It is set up when the cell ANY DISK FULL becomes true Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_MSCP_IO_RATEDEV_F_MSCPOP This contains a value representing the average number of MSCP served I O requests per second to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_MSCP_IO_RATE_TALLY Derived This contains the sum of the values representing the average MSCP served I O rate for the current disk subrecords which were selected by the most recent DISK_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_MSCP_PAGING_IO_RATEDEV_F_MSCPPG This contains a value representing the average number of MSCP served paging I O requests per second to and from the c
517. s mutually exclusive with BEGINNING and ENDING ENDING date Specifies the ending time for the data selection By default 23 59 or NOW is used FILTER keyword The FILTER qualifier allows you select a subset of the daily or history data for interactive displays Process data and disk data can be filtered Hotfile data is also filtered When you specify filtering by process a hotfile record is selected if accessed by the specified process When you specify filtering by disk device a hotfile record is selected if located on the specified device For hotfile records matching both process and disk device specify filtering by both process and device 242 Performance Manager Administrator Guide SELECT Process data can be filtered by using any of the filter keywords USERNAMES IMAGENAMES PROCESSNAMES ACCOUNTNAMES UICS PIDS or WORKLOADNAMES If a process record s identification information matches any of the identification specifications that are specified then that record is selected Likewise disk data can be filtered by using any of the filter keywords VOLUMENAMES and DEVICENAMES If a device record s identification information matches any of the volume names or device names that are specified then that record is selected Keyword Description FILTER USERNAMES string Specify FILTER USERNAMES to select all process records with the username matching any of the specified strings FILTER IMAGENAMES string
518. s pattern color click the Foreground or Background button Performance Manager displays a color mixing dialog 4 Click Help for information on how to use the color mixing dialog Modify Parts To modify the parts of an instrument 1 Pull down the Edit menu choose the Build Mode menu item and click the Enable submenu item 2 Select the instrument by clicking on it 3 Pull down the Edit menu choose the Modify menu item and click the Parts menu item in the submenu Performance Manager displays the appropriate Parts dialog for the selected graph or chart in one of the following boxes m Bar Graph Parts Definition dialog m Strip Chart Parts Definition dialog Meter Parts Definition dialog The Bar Graph Parts Definition dialog lets you specify m Title String Set the toggle button to display a title Enter a title in the text entry box m Data Name Set the toggle button to specify the title to be the metric name overriding any supplied string 352 Performance Manager Administrator Guide Bar Graph Parts How You Edit the Panel Instruments W Title string CJ Wi Data Name Data Labels Tags Ticks s W Meter Fields s E E Short Ticks O Units Label Orientation Tags Ticks List of Instances Type of One Metric Left g Right o Nbr of List Entries 7 Top O Bottom O Title Font Click the Title Font button and Performance Manager displays the Font Selection dialog as s
519. s the minimum working set extent for a process created by the Create Process CREPRC system service or the DCL command RUN Process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN PQL MWSQUOTAPAR F PQL MWSQUOTA The value of the SYSGEN parameter PQL MWSQUOTA which sets the minimum working set quota for a process created by the Create Process CREPRC system service or the DCL command RUN Process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN QUANTUMPAR F QUANTUM The value of the SYSGEN parameter QUANTUM which defines the maximum amount of processor time a process can receive before control passes to another process of equal priority that is ready to compute for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN RESHASHTBLPAR F RESHASHTBL The value of the SYSGEN parameter RESHASHTBL which defines the number of entries in the lock management resource name hash table for the local node for the current interval record in LOCAL domain and for the
520. s well as the time period of interest can be specified The data can be averaged into a series of records representing whatever time interval the user wishes as long as it is a multiple of the collection interval For instance data collected every two minutes can be averaged into half hour records You can select a specific set of items disks or processes for output An ASCII or binary file can be created Either daily or archived data can be exported DECwindows Interface Performance Manager includes DECwindows Motif Interfaces for real time displays and for analysis and graphing These interfaces are installed only if the necessary windowing libraries are found in SYSSLIBRARY and the necessary DECwindows directories are accessible Chapter 1 Introduction 17 What to Expect from Performance Manager What to Expect from Performance Manager Performance Manager analyzes collected system data required for determining whether a specific resource is causing a performance problem The data collected at your site may cause Performance Manager rules to fire To fire means that when all the data has been processed Performance Manager examines the count of a rule s occurrences and if there are enough occurrences for a particular rule that rule is said to fire This causes an entry for that rule to be placed in the report file in the form of a conclusion However you can modify or disable Performance Manager rules or threshold values which gove
521. screen The following code is sample output using TRACE RULETRACE of UMO10 condition 1 is TRUE 9 RULETRACE of UMO10 condition 2 is FALSE Chapter 5 Customize the Knowledge Base 151 Components of Rules The Performance Manager displays this debugging information for each rule condition when the condition is evaluated and terminates with the first FALSE condition It is possible that the Performance Manager does not output a trace for your rule This occurs when another rule with a rule condition identical to one of your rule s conditions is evaluated to FALSE In this case the Performance Manager does not evaluate your rule because it is known to be false The relationship between domains and data cells is fixed Rules are in domains Data cells are in one or more domains m Rules defined in the CLUSTER or SUMMARY domain can reference data cells only in the CLUSTER or SUMMARY domain respectively m Rules defined in the LOCAL domain can reference data cells in the LOCAL domain and in other domains but only if there is an index specifier for the specific domain m Rules defined in any of the remaining seven domains COMMUNICATION CONFIGURATION CPU DISK FILE PROCESS or TAPE can reference data cells in their own domains or the LOCAL domain The rule can reference data cells in any other six domains only if there is an index specifier for the other domains Rule Condition Element A rule must have one or more rule conditio
522. sday 04 JAN 2006 11 00 to 11 45 VAXMAN OSF PROJECT PLAN TXT 5 40672 3 0 VAXMAN REMINDER WRK DIR 1 LANGELO CIRRUS ROM DIR 1 Non Virtual QIO DECPS DATABASE PSDC YQUEM 19910CT04 CPD 1 DECPS DATABASE PSDC SCHEDULE DAT 1 DECPS DATABASE DIR 1 The following statements are keyed to the columns in the previous example 1 The disk device and volume name 2 The average I O rate in I O operations per second for the intervals of time when the file is hot Chapter 3 Evaluate Performance in Detail 61 Performance Evaluation Report 3 The peak I O rate in I O operations per second for the interval record when the file is hottest 4 The percentage of I O activity devoted to file READs 5 The percentage of I O activity where a split I O operation occurred 6 The interval time for the peak file activity The hours and minutes are preceded by the day of the month In this case 4 11 46 represents October 4 at 11 46 a m Because this report can span multiple days the Performance Manager reports the day as well as the time in this field 7 The record count indicates the number of Performance data records during the reporting period in which the file is hot 8 The file specification for the hot file If the file is deleted before the Performance Manager detects its specification the FID is provided in parentheses instead of its name All non virtual QIO activity to the disk is reported under the filespec N
523. se of this qualifier ADVISE PERFORMANCE EXPORT CLASS NODEFAULT PROCESS NODE MYNODE 5 BEGIN time END time _ PROCESS STATISTICS PRIMARY KEY IMAGENAME SECONDARY KEY PID _ CLASSIFY BY USERGROUP MODEL TRANSACTIONS This command produces an output file called PSPASDUMP DAT in the current directory in ASCII CSV format The process data records contain all detail fields including the workload name with which it is associated sorted by imagename first then PID within that The CLASSIFY BY qualifier is needed to specify the Family definition and Workload definitions that define the workload groups SCHEDULE day hour range NOday Where day is one of the following words m EVERYDAY m WEEKDAYS m WEEKENDS m MONDAY m TUESDAY m WEDNESDAY m THURSDAY m FRIDAY m SATURDAY m SUNDAY The hour range is in the form of m n where m is an integer hour from 0 to 23 and n is an integer hour from 1 to 24 and larger than m Use the SCHEDULE qualifier to select a sub set of the performance data for exporting By default all data between the BEGINNING time and the ENDING time is selected Use the day keywords with hour ranges to specify what data is to be included Negate any of the day keywords to omit data for a range of days Do not specify any hour ranges with negated keywords Chapter 6 Performance Manager Commands 189 ADVISE PERFORMANCE EXPORT Examples TYPE ASCII BINARY There are t
524. select any of the following class items Keyword Meaning ALL Report all optional class statistics along with default classes CPU Report CPU statistics DEFAULT STATISTICS Report CPU DISK IO MEMORY PAGE FAULT and XQP CACHE statistics DEVICE Report device statistics DISK Report disk statistics IO Report I O statistics 182 Performance Manager Administrator Guide ADVISE PERFORMANCE EXPORT Keyword Meaning LOCK Report lock statistics MEMORY Report memory statistics PAGE FAULT Report page fault statistics PROCESS NO EXTENDED Report process statistics NO IMAGE ALL SYSTEM COMMUNICATION Report system communication services statistics XQP CACHE Report XQP statistics You can negate these keywords to indicate that reporting of a particular class of data is not wanted Also you can specify CLASS ALL negated keyword to allow an all but these capability The DEFAULT STATISTICS class of statistics is always on unless specifically negated For example if CLASS LOCK is specified then CPU DISK IO MEMORY PAGE FAULT XQP CACHE and LOCK statistics are exported To disable any classes of data which are part of the DEFAULT STATISTICS group you must specify NODEFAULT STATISTICS Therefore if you want to report only disk and CPU data specify CLASS NODEFAULT STATISTICS CPU DISK The PROCESS keyword allows you to specify the optional keywords EXTENDED IMAGE or ALL The presence of the EXTENDED keyword in
525. sis O Full Analysis O Tabular Report O Configuration Section O Disk Section L Summary Section O Process section O Performance Report Ll Process Statistics L1 Disk Statistics O Pool Statistics L Summary Statistics O Lock Statistics O Mode Statistics L Tape Statistics O CI Statistics L SCS Statistics O Rule Statistics O Hot File Statistics Prepare Process Statistics A By Node wv Cluster wide wv Cluster wide with By Node detail Cancel 2 Enter a file name and select report sections you want to save 3 Click OK to apply your selections and save the indicated reports The dialog box is removed from the screen and the file created 4 Click Reset to redisplay the default settings if you changed settings without applying them OR Click Cancel to dismiss the dialog box without changing any settings or saving a report 262 Performance Manager Administrator Guide How You Control the DECwindows Interface Monitor the Work in Progress To monitor the work in progress 1 To display the Performance Manager Work in Progress dialog box release on Work in Progress Selecting Data 100 Complete Reading data for node YQUEM Reading data for node GALLO The work that is being monitored is the reading of the Performance Manager data files and the building of the internal data structures needed for the requested analysis This work is started as a result of data selection See the section How Y
526. sor Mode Procesor e 191 Terminal Input Images e 191 Terminal Input Users e 191 Terminal Input Workloads e 191 Terminal Thruput Images e 191 Terminal Thruput Users e 191 Terminal Thruput Workloads e 191 Throughput Disk Device e 191 Throughput Disk Volume e 191 Throughput Files e 191 Throughput Images e 191 Throughput Users e 191 Throughput Workloads e 191 User Disk Operations e 191 User Mode Procesor e 191 User Volume IO e 191 Index 581 VA Space Images e 191 VA Space Users e 191 VA Space Workloads e 191 Workload Disk Operations e 191 Write Operations Disk Device e 191 Write Operations Disk Volume e 191 Write Operations Files e 191 WS Size Images e 191 WS Size Users e 191 WS Size Workloads e 191 Graphs components of e 111 117 286 Custom e 126 295 display of e 286 293 legend e 111 selecting multiple e 108 selection of e 274 stacked e 288 top system use e 294 Gvalid faults e 79 H Hard fault rate panels e 330 faults e 79 388 page fault in performance evaluation report e 34 HEIGHT keyword for generating graphs e 191 High threshold bar graph setting of e 347 strip chart setting of e 348 Histogram e 64 70 asterisk 4 e 65 CPU utilization e 64 plot resident time for image system load and performance e 64 provide finer resolution e 64 History data how to select e 191 file descriptors archive schedule e 317 create e 317 create workload classifications e 317 data reduction e 317 deleting e 319 modify e 31
527. splay The following illustration is an example of the I O display Chapter 11 Use the Character Cell Real time Display 399 Evaluate Performance Using the Investigate Displays Isolate the Cause of a CPU Limitation If an examination of the system overview reveals a CPU limitation you can investigate the cause of the limitation in more detail using the CPU display The following illustration is an example of the CPU display Examine the following CPU display indicators Any available CPU Look at the Idl bar CPU Idle time in the CPU mode bars in the middle of the CPU display If there is no idle time the CPU is a bottleneck A few processes blocking other processes The blocking high priority process might be running an inefficient program acting as a server or acting as a process with which other processes must communicate Look at the Priority bars for high priority active processes or at the Process bars for high priority processes with a high CPU time percentage Corrective action might include changing process priorities in the user authorization file defining priorities in the user login command file or changing the priorities of processes while they execute Lost CPU time Look at Page wait Swap wait and Pg Swp Wt CPU time might be lost because the CPU has to wait for disk transfers or page or swap I O to complete A high value of Pg Swp Wt is cause for concern It indicates a memory problem resulti
528. ssive operations rate on a disk on the local node for the current interval It is set up when the cell ANY DISK OVER THRESHOLD becomes true Data Type INDEX Appendix C Performance Manager Data Cells 439 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_PAGING_IO_RATEDEV_F_PAGOP This contains a value representing the average number of paging I O requests per second to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_PAGING_IO_RATE_TALLY Derived This contains the sum of the values representing the average paging I O rate for the current disk subrecords which were selected by the most recent DISK_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_PAGING_THRUPUT_RATEDEV_F_PAGIO This contains a value representing the average number of bytes per second for paging I Os transferred to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_PAGING_THRUPUT_TALLY Derived This contains the sum of the values representing the average paging throughput rate for the current disk subrecords which were selected by the most recent DISK_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK 440 Performance Manager Administrator Guide Performance Manager Data Cell
529. stem metrics you can select see Appendix C Only numeric metrics in Domain LOCAL can be requested Use the multi node keypad to perform the following tasks m Get Help m Zoom in on node m Change to the next node m Change to the previous node m Change to the resource display m Change to the next metric m Change to the investigate display m Change to the previous metric The following illustration shows the functions of the multi node display keypad Help Resource Invest KEYPAD DEFINITION Key Display igate Display for the Zoom MULTI NODE DISPLAY in on Node Z Exit PSPA RT Chapter 11 Use the Character Cell Real time Display 377 Display Single Node Statistics Display Single Node Statistics To navigate through the list of nodes 1 Press KP2 or KP5 The arrow identifies the node you choose to monitor 2 Press the zoom in on node key KP7 in the multi node keypad Performance Manager displays a single node screen DE Cterm 1 File Edit Commands Options Print The Performance Manager single node screen provides the following three sections m CPU utilization and mode statistics m Top processes statistics m Top device statistics The Process Statistics and Device Statistics sections can be updated with a new set of metrics The title of the currently selected section is in reverse video Press KP3 to select a different section 378 Performance Manager Administrator Guide Display Single Node
530. ster wide Tape Statistics ss 58 Interpret cluster wide Hot File Statistics e 61 Interpret cluster wide Summary Statistics ses 62 HistogramS ssh hh ha 64 Image Residence Histograms lla ll 64 Contents 5 Tabular Report Secti0ns aa AA he hh ah hehehe 71 System Configuration Data 2 72 Summary Statistics SecLloris ssec restes Naa ceeded needa cue ka ipu ted ecu uia 74 System Communication Service Rates eect teen eee ee eee 85 Disk and Server Statistics Section 000 000 ccc sisse 88 Process Metrics Data 0 ccc tnt nent ha 90 Cluster Summary Statistics with By Node Breakout 22 0000 c cece eee eee 93 Cluster Disk and Server Statistics with By Node Breakout 2 22 ee eee 99 Chapter 4 Generate Historical Graphs 107 Generate Predefined Graphs 0 000 c ccc sls 107 Generate Graphs from the DCL Level 000 0 c eect nets 108 Generate Graphs in Command Mode 00000 c eect t eet teens 108 Generate Multiple Graphs 0 a 111 Components of Graphs scissa sodas PPP PG qUR3 Pd a E PPERPRPTQPG qud qe ue P MY Epp 111 Composite Graphs dama le plu e ep eu Vau lu ple a epis a e e v xu d a a d 113 Stacked Graphs i 2i4sesdacad d du GEAR AY TEEPEE TEE P O PG ORPGQGTG 344 343 334 P3 GG PES 113 Create Typical Time Period Graphs 0 0 20 cece sss nn 113 Sched
531. ster wide displays e 386 Collection definitions create e 297 delete e 297 modify e 297 Index 577 COLOR keyword e 117 Colors and patterns Change e 339 Command mode for interactive displays e 239 Compat e 77 Compute Queue graph type e 191 Conclusions e 15 18 236 Analysis report e 235 definition e 156 order and type in analysis report e 21 Copy panel instruments e 342 Coverage lists device name e 298 process name e 298 CPU ID e 77 Modes graph e 191 modes panel e 326 queue e 76 queue panel e 329 Utilization graph e 108 191 Utilization histogram e 64 utilization panels e 325 create a CSV file e 559 create a graph in Excel e 562 Custom graphs e 126 191 295 creating e 126 default title e 111 example e 127 metrics e 126 prompt mode e 126 pie charts e 232 295 Customize menu e 306 D Daily data e 268 Data collection schedule e 298 file minimizing file reads e 108 Data cell as evidence e 154 associated with domains e 150 availability to rules e 150 effects of thresholds e 137 index specifiers e 156 list of predefined thresholds e 137 list of types e 156 type 578 Performance Manager Administrator Guide Boolean e 158 index specifier e 162 numeric e 159 scan routine e 159 string e 159 tally e 160 time e 159 types and usage e 156 Datagram e 85 Deadlock e 84 Debugging user written rules e 150 DEC CRT e 373 DECNET graph type e 191 DECnet Phase V support node name translation e 406 DECwindows interface e 1
532. stomize e 306 locked e 259 306 PSDC PARAMS DAT e 306 release e 306 unlocked e 306 view e 306 when changes become effective e 239 PATTERN keyword e 117 Patterns and colors bar graph set e 350 set e 349 strip chart set e 351 Peak hold units bar graph setting of e 347 Performance activity report navigation of e 282 Agent 584 Performance Manager Administrator Guide collection schedule e 298 delete after e 298 disk space needed e 298 end date e 298 hot file queue e 298 node definition list e 298 setting parameters e 298 start date e 298 view settings e 296 work set quota e 298 evaluation report e 15 62 266 CI NI and adapter statistics e 55 CPU mode statistics e 51 cpu statistics e 44 disk statistics e 56 hot file statistics e 61 I O statistics e 46 interpreting e 34 62 lock traffic e 54 memory statistics e 48 navigation of e 280 pool statistics e 50 process statistics e 38 40 SCS message and transfer rates e 52 selection of e 274 summary statistics e 62 tape statistics e 58 evaluation window e 280 Manager reporting schedule e 269 Performance Agent status e 259 Performance analysis data classification of e 268 source of e 268 display reports e 278 select analysis e 274 dump reports e 274 graphs e 274 last hour e 277 performance e 274 report options e 274 tabular final reports e 274 tabular interval reports e 274 select data default 266 specific e 267 select nodes e 274 set reporting period e 269
533. string Specify FILTER IMAGENAMES to graph all process records with the imagename matching any of the specified strings Do not specify any trailing EXE nor the file version device or directory PROCESSNAMES string Specify FILTER PROCESSNAMES to graph all process records with the processname matching any of the specified strings The match string is case sensitive so if the process names have any lowercase letters spaces or tabs use double quotes when you enter the value e g FILTER PROCESSNAMES RTserver ACCOUNTNAMES string Specify FILTER ACCOUNTNAMES to graph all process records with the accountname matching any of the specified strings 194 Performance Manager Administrator Guide Keyword WORKLOADNAMES workloadname ADVISE PERFORMANCE GRAPH Description Specify FILTER WORKLOADNAMES to graph all process records associated with any of the specified workloads This filter is valid only if the CLASSIFY_BY qualifier is used to specify a classification scheme for your workload data UICS uic Specify FILTER UICS to graph all process records with the UIC matching any of the specified UICs An asterisk may be used to wildcard either the group or user field of the specified UICs PIDS pid Specify FILTER PIDS to graph all process records with the PID matching any of the specified PIDs VOLUMENAMES string Specify FILTE
534. struments To edit instruments wm Pull down the Edit menu on any open panel and release on the menu item you want The Edit menu lets you perform the following actions m Modify instruments m Enable build mode m Create instruments Strip charts A graphical representation of a data item in an X Y wide grid New data items enter from the right and the entire chart moves to the left with the oldest data disappearing at the left edge A strip chart gives a pictorial view of historical data An example of a strip chart is a medical electrocardiogram EKG Chapter 10 Customize the DECwindows Motif Real time Display 345 How You Edit the Panel Instruments Bar graphs A graphical representation of a data item in which the height or length of the bar represents the magnitude of the data item At each interval the bar is redrawn for the new value of the data item so the bar grows or shrinks as intervals progress Meters The numerical or string representation of the data item An automobile odometer is an example of a meter Copy instruments Delete instruments m Assign metrics Assign launch panels To edit instruments 1 Enable build mode 2 Select an instrument 3 Select an editing function Enable the Build Mode Enabling the build mode lets you create and edit instruments within the displayed panel To enable the build mode 1 Pull down the Edit menu 2 Click the Build Mode menu item 3 Click the Ena
535. sts are New ENQ The number of new locks requested enqueued per second Converted ENQ The number of lock conversion requests per second DEQ he number of locks released dequeued per second Blocking AST The number of blocking ASTs received per second Use of blocking ASTs allows a process to lock a resource and then release it only when another process requests that resource When another process requests a lock on the resource a blocking AST is delivered to the process currently holding the lock Directory Functions The number of messages per second for directory operations There are three categories the rate for lookups in a directory the rate for inserts in a directory and the rate for deletes from a directory Deadlock Messages The number of messages per second required for deadlock detection 84 Performance Manager Administrator Guide Lock Counts Tabular Report Sections This section reports lock statistics With the exception of the Total Resrcs field all data is collected from OpenVMS performance statistics ENQ Wait The number of times lock requests were forced to wait ENQ NotQD The number of times a lock request was not granted process failed to get lock and did not wait DLock Search The number of times a search for deadlocks was initiated by the system The system parameter DEADLOCK WAIT defines the number of seconds that a lock request must wait before the system initiates a deadlock search
536. summaries The GRAPHS option lets you control how data is to be selected for subsequent graphing operations including pie charts The GRAPHS option has the following eight sub options IMAGENAMES m USERNAMES m HOTFILES m USERVOLUMES m IO DEVICES and workloads m BY NODE Chapter 4 Generate Historical Graphs 109 Generate Predefined Graphs m ALL m DEFAULT DEFAULT generates graphs for IMAGENAMES USERNAMES and IO DEVICES and workloads DEFAULT is used in the absence of any specified graph processing options ALL is equivalent to the complete list IMAGENAMES USERNAMES HOTFILES USERVOLUMES IO DEVICES and workloads and BY NODE Note that when you are processing a large amount of data each option can pose significant additional CPU and memory demands on your process If you specify NOALL only system level metrics are saved NOALL is helpful when you need to select data as fast as possible and retain the ability to generate graphs for the system level metrics The system level metrics are always saved BY NODE as well The processing options IMAGENAMES USERNAMES HOTFILES IO DEVICES and workloads and USERVOLUMES cause the selection process to maintain the graph statistics for each unique occurrence of an image name user name and so forth The BY NODE option causes these statistics to be maintained on both a per node and a composite all nodes basis The BY NODE option can increase the memory demands to s
537. t Plots the five plus Other HSC channels that experienced the largest throughput rate in Kilobytes per second The metric graphed is CHANNEL THRUPUT Note The channel names are provided in the format nodename n where n represents the channel number K SDI on the HSC node indicated by node name If the channel cannot be identified the character u is substituted for n See logical name PSDC hscname hscunitnumber in the Performance Agent Administrator Guide TYPE TOP CLUSTER RULE OCC Top Cluster Rule Occurrences Plots the five plus Other rule identifiers that fired as a rate per hour The metric graphed is CLUSTER OCCURRENCES and is available only from history data 220 Performance Manager Administrator Guide ADVISE PERFORMANCE GRAPH TYPE TOP COMPAT PROCESSOR Top Compat Mode Processor Plots the five plus Other processors in terms of time spent in compatibility mode as a percent of CPU time The metric graphed is P COMPAT TYPE TOP CPU IMAGES Top CPU Images Plots the top five plus Other Images consumers of CPU time by image name The metric graphed is CPUTIME TYPE TOP CPU RULE OCC Top CPU Rule Occurrences Plots the five plus Other CPU rule identifiers that fired as a rate per hour The metric graphed is CPU OCCURRENCES and is available only from history data TYPE TOP CPU USERS Top CPU Users Plots the top five plus Other Users consumers of CPU time by user name
538. t Panel CPU Utilization CPU utilization for this Percentage with history CPU Utilization current Direct IO Rate Disk IO Rate Disk Thruput Hard Fault Rate process CPU utilization for this process Direct I O for this process Disk I O for this process Disk throughput Hard page fault rate 10096 maximum Percentage 10096 maximum Operations per second Operations per second Kilobytes per second Faults per second Chapter 9 Use the DECwindows Motif Real time Display 327 Default Panel Descriptions Label Soft Fault Rate Buf I O Rate Wssize Wsdefault Wsquota Wsext Private Global Virtual Addr User Name Account Process ID Image Name Process State Process Priority Base Priority Metric Soft page fault rate Buffered I O rate Working Set size for this process Working set default setting for this process Working set quota setting for this process Working set extent setting for this process Private memory size for this process Global memory size for this process Virtual address size for this process User name for this process User account for this process Process ID PID for this process Current image name for this process Current process state Current priority for this process Base priority for this process Units Faults per second Operations per second Pages Pages Pages Pages Pages
539. t This value is computed as Average device queue length Rate Queue length The average number of disk I O requests waiting for service Space Used The percentage of the total disk volume space that is allocated Reported only for mounted disks Server By Cluster Format The following example shows server statistics in BYCLUSTER format In the By Cluster format for server statistics the leftmost column of the tabular report contains the name of each server that is accessible cluster wide Server names are followed in parentheses by the name of the CPU or HSC server with which they are associated Each server name begins a line of server statistics Each statistic shows the activity for the specific server due to all nodes in the cluster as shown in the following example CLUSTER Server Statistics Alloc Paging Swping 6 Chapter 3 Evaluate Performance in Detail 105 Tabular Report Sections Server By Node Format Work Avail VAX VAX VAX VAX VAX VAX VAX VAX In the By Node format for server statistics the leftmost column of the tabular report contains the name of each server that is accessible cluster wide This is followed by the name of each node from which the server was accessed Each server by node name begins a line of server statistics as in the case of By Cluster format These statistics show the contribution of each individual node to the activity of a specific server Server names are follow
540. t avg The number of process inswaps performed during the last example interval The following CPU statistics are reported in By Cluster and By Node formats 96 Performance Manager Administrator Guide Tabular Report Sections CPU Busy Percentage of time the CPU time spent in interrupt stack busy wait kernel executive supervisor user and compatibility modes This is the sum of System CPU and Task CPU 96 System CPU The sum of the interrupt kernel executive and busy wait CPU percentages Task CPU 9 The sum of the supervisor user and compatibility mode busy percentages CPU Queue avg The average number of processes waiting for the CPU There is a queue if this number is greater than one 1 the process that would have run in absence of the collection process Equivalent to the sum of processes in the computable queue COM CPU Idle The percentage of time that the CPU was idle CPU IO Idle 9 7 The percentage of time that the CPU and all disk devices selected for data collection were idle CPU Only The percentage of time non overlapped CPU time that the CPU was busy and no disk device selected for data collection was busy I O Only The percentage of time non overlapped I O time that the CPU was idle and at least one disk device selected for data collection was busy Multi I O The percentage of time that two 2 or more disk devices selected for data co
541. t Graphs and Pie Charts In the Maximum X POINTS Graph the value of X POINTS is 480 Therefore the graphing facility did not average the data This following command generates the following graph ADVISE PERFORMANCE GRAPH BEGINNING 19 FEB 2006 04 00 ENDING 19 FEB 2006 20 00 _ TYPE CPU UTILIZATION NODE GALLO OUTPUT CHACPU XP 480 PS _ FORMAT POSTSCRIPT CHARACTERISTIC PATTERN X POINTS 480 CPU UTILIZATION tois GALLO Percent of CPU SINIY A FAILS H oe EES latatkaM Jepack ttt In the Minimum X POINTS Graph the value of X POINTS is 8 The graphing facility averages every 15 data points in the four hour time span Chapter 4 Generate Historical Graphs 125 Generate Custom Graphs This command generates the graph ADVISE PERFORMANCE GRAPH BEGINNING 19 FEB 2006 13 00 ENDING 19 FEB 2006 17 00 TYPE CPU UTILIZATION NODE GALLO OUTPUT CHA4CPU XP 8 PS _ FORMAT POSTSCRIPT CHARACTERISTIC PATTERN X POINTS 8 CPU UTILIZATION ALL Percca of CPU Sar A QUAS Generate Custom Graphs The Custom graph type behaves differently than all other graph types because to control the graph data you must specify one metric and up to six data items or up to six metrics and one data item You can use metrics from one of the following groups for a custom graph m System metrics m Process metrics selected by user name image name or workload name m Disk device metrics selected by device nam
542. t I O Transaction Transaction class The schedule file is a file PSDC SCHEDULE DAT that resides in the PSDC DATABASE area and controls when Performance Manager daily data is recorded A shadow set is one or more compatible physical disk volumes connected together for volume shadowing and represented by a virtual unit Thus the term shadow set refers to the physical unit and the virtual unit Each time a process references a virtual page that is not in its working set but is in memory a soft page fault is generated Split I O is the number of additional physical disk I O operations required to complete a single user s I O request which could not be satisfied in a single I O to a device A transaction is a quantifiable unit of work that typically delineates a single processing step in computer systems A transaction class is a group of related transactions They may be related by the function they perform by the users who initiate them or by other quantities you define Transactions may also be determined by the system resource demands The Performance Manager software generally refers to a transaction class as a workload Transaction class workload A transaction class workload is a workload that contains data bucketed by a workload definition defined in terms of images Transaction workload family User defined graph A transaction workload family is a set of image based workload definitions Usergroup workload family
543. t of all Performance Manager thresholds is presented in the Performance Manager Thresholds table at the beginning of this chapter To implement a change like this is relatively straightforward Simply add the following line to your MYRULES VPR file Threshold TD DISK QL MAX 0 66 EndThreshold Assume that you would like to make this change plus raise the threshold of free space remaining on a disk from 5 percent to 10 percent for use in the rule that you just finished adding in the last section To change more than one threshold you might want to change your format within MYRULES VPR to the following for greater clarity Threshold TD DISK QL MAX 0 66 TD MIN DSKSPC PCT 0 10 EndThreshold There are other cases when a threshold might need to be changed For example if you own older RF31 disk drives you should change the RF31 threshold values as they now reflect the performance of the newer RF31T disk The internal model number used by OpenVMS to identify disk types is no longer unique this is the first case of an ID number being re used for a newer disk but more recycling of IDs is expected in the future Also certain disks may be able to process many more I Os per second than indicated by the given disk thresholds if they are able to make effective use of their embedded or HSC based disk cache If you have many I O rule firings but upon investigation find relatively low queue lengths on disks that are processing I Os at
544. t packets for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LRPCOUNTVPAR F LRPCOUNTV The value of the SYSGEN parameter LRPCOUNTV which establishes the upper limit to which LRPCOUNT can be automatically increased by the system for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN LRPSIZEPAR F LRPSIZE The value of the SYSGEN parameter LRPSIZE which indicates the size in bytes of the large request packets for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MAXPROCESSCNTPAR F MAXPROCESSCNT The value of the SYSGEN parameter MAXPROCESSCNT which is the maximum number of processes allowed on the system for the current interval for the local node Data Type NUMERIC Appendix C Performance Manager Data Cells 509 Performance Manager Data Cells Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN MINWSCNTPAR F MINWSCNT The value of the SYSGEN parameter MINWSCNT which is the minimum working set size the minimum number of fluid pages not locked in
545. tagram messages received over the connection The metric graphed is MGS R TYPE TOP MGS S Top Messages Sent Plots five nodes plus Other in terms of number of application datagram messages sent over the connection The metric graphed is MGS S TYPE TOP MP SYNCH PROCESSOR Top MP Synch Mode Processor Plots the five plus Other processors in terms of time spent in MP synchronization mode as a percent of CPU time The metric graphed is P MP SYNCH TYPE TOP MSCPIO FILES Top MSCP I O Operations Files Plots the five plus Other files that incurred the highest MSCP I O rates The metric graphed is FILE MSCPIO TYPE TOP PAGING DISKS Top PG amp SWP Operations Disk Device Plots the five plus Other Disks disk devices that incurred the highest I O paging and swapping rates The metric graphed is PAGIO TYPE TOP PAGING FILES Top PG amp SWP Operations Files Plots the five plus Other files that incurred the highest I O paging and swapping rates The metric graphed is FILE PAGIO TYPE TOP PAGING VOLUMES Top PG amp SWP Operations Disk Volume Plots the five plus Other Volumes disk volumes that incurred the highest I O paging and swapping rates The metric graphed is PAGIO TYPE TOP POOL RULE OCC Top Pool Rule Occurrences Plots the five plus Other pool rule identifiers that fired as a rate per hour The metric graphed is POOL OCCURRENCES and is available only from history data
546. tains a sample daily command procedure PSPASEXAMPLES PSPASDAILY COM that you can use as a template You can use the OpenVMS Mail Utility to send the brief analysis to you directly Real time Displays of Performance Data The real time feature provides continuously updated displays of data You can produce and customize three dimensional bar graphs strip charts and digital meters to evaluate and investigate performance data using Performance Manager Motif real time display 16 Performance Manager Administrator Guide DECwindows Interface You can also produce tabular and graphical displays to evaluate and investigate performance data by using the Character Cell interface with optional ReGIS graphics Using Playback mode you can view displays of data recorded earlier in either a continuous flow or step mode Graphing Facility To graph chronologically any group of metrics stored in the database use the ADVISE PERFORMANCE GRAPH command to produce a wide range of predefined graphs You can also define your own custom graphs if the predefined graphs do not meet your specific needs You can select a specific set of items disks or processes for graphing For more information on graphing see the chapter Generate Historical Graphs see page 107 Data EXPORT Facility The performance data from Performance Manager files can be written out to a file for subsequent processing with third party software tools The desired classes of data a
547. tains the number of local blocking ASTs queued per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LOCAL_LOCK_CONVERSION_RATEMET_F_ENQCVT_LOC This contains the number of new local enqueue lock conversion requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LOCAL_LOCK_DEQUEUE_RATEMET_F_DEQ_LOC This contains the number of local dequeue lock requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LOCAL_LOCK_ENQUEUE_RATEMET_F_ENQNEW_LOC This contains the number of new local enque requests per second for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP 464 Performance Manager Administrator Guide Performance Manager Data Cells LOCKIDS IN USEMET F LOCK CNT This contains the number lock IDs for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LOCKID_TABLE_SIZEMET_F_LOCK_MAX This contains the lock ID table length in entries for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP LOCK_RESOURCES_IN_USEMET_F_RESOURCE_CNT This contains the number lock resources known by the local node for the current interval Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PR
548. tal number of global pages that can be created for the current interval for the local node Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP 506 Performance Manager Administrator Guide Performance Manager Data Cells SYSGEN GBLSECTIONPSPAR F GBLSECTIONS The value of the SYSGEN parameter GBLSECTIONS which sets the number of global section descriptors allocated in the system header at bootstrap time for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN GROWLIMPAR F GROWLIM The value of the SYSGEN parameter GROWLIM which sets the number of pages that the system must have on the free page list so that a process can add a page to its working set when it is above quota for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN IOTAPAR F IOTA The value of the SYSGEN parameter IOTA which sets the I O time allowance in 10 millisecond units used to charge the current residence quantum for each voluntary wait for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM T
549. tatistics section of the Performance Evaluation Report use the following qualifier INCLUDE POOL STATISTICS To disable the pool statistics display from the report use this qualifier INCLUDE NOPOOL STATISTICS The following example shows the pool statistics of the Performance Evaluation Report Performance SUPPLY VAX 11 78 Evaluation Wednesday 14 JAN 1997 00 00 to 12 16 The following table gives the average pool resources used and allocated on this node N A means not applicable SRP NP POOL Avg number in use 499518 Max number in use 2122 516976 Number of intvls w expansns 0 0 Allocation XRPCOUNT 1249 799744 Virtual Alloc XRPCOUNTV 160 4996 2399744 50 Performance Manager Administrator Guide Performance Evaluation Report The following statements are keyed to the columns in the previous example Large request packets Intermediate request packets Small request packets 1 2 3 4 Nonpaged pool in bytes 5 Number of locks 6 Number of named resources known by the Distributed Lock Manager For OpenVMS Versions 6 0 and higher the metrics for LRP IRP and SRP are obsolete The following sample shows an example of this report Performance MUMMS VAXstation 3100 GPX Page 1 Evaluation PA Vx x Saturday 19 MAR 2006 23 00 to 23 59 The following table gives the average pool resources used and allocated on this node N A means not applicable Avg number in use 2449519 Max number in use 2456128 Num
550. te reports and graphs for all disk records with the volumename matching any of the specified strings Do not specify any trailing colon Devicenames Specify a list of strings separated by commas spaces or tabs to generate reports and graphs for all disk records with the devicename matching any of the specified strings Do not specify any trailing colon Process Your Options You can enter any of the following options Analysis Directs the Performance Manager to analyze the data to generate an Analysis Report Performance Directs the Performance Manager to analyze the data to generate Performance Evaluation and Process Statistics Reports The primary key for presenting process statistics defaults to Interactive Batch Network Detached or Other organization The secondary key defaults to Imagename 274 Performance Manager Administrator Guide How You Select Data for Analysis While viewing the resulting report an option is provided to re sort the process statistics Dumps Specifies that the unanalyzed data will be made available for user inspection Since no pre processing of the data is required for dumps no overhead is added to the analysis process Refer to the CA Performance Manager for OpenVMS Administrator Guide for more information about dump reports Tabular Final Directs the Performance Manager to analyze data to generate the Tabular Report sections averaged over the entire analysis period Report Option
551. tents 5 Chapter 1 Introduction 13 CA Performance Management for OpenVMS 2 2 2222 2 sse 13 Performance Manager 0 AA p acr pa a a rh a a a a a ya hat 14 Performance Manager Features 22 ls rn 15 Knowledge Base and Rules Compiler 2 see 15 Analysis and Reporting Facility sse 15 Real time Displays of Performance Data see 16 Graphing Facility 2 osorereliescuererer ee Py DILIS diedoddsatadgudtadeneseddeedasaddiadoas 17 Data EXPORT Facility rn re 17 DECwindows Interface c esee RR ERR RR KANG ERE RA EE nG 17 What to Expect from Performance Manager 2 2 2222 se 18 Cross Platform Support sls rr 19 Chapter 2 Analyze Performance 21 Analy SiS REDO Sennan e See a ee LIMINE UU SS REL 21 Interpret the Analysis Reports aa cece eee cere GA hus 22 Brief Analysis Reports sss nn 30 Interpret the Brief Analysis Report rr 30 Chapter 3 Evaluate Performance in Detail 33 Performance Evaluation Report a AG ENE ENES ANNA Naa 33 Interpret the Process Statistics sel 34 Interpret Pool Statistics AA ls lll 50 Interpret CPU Mode Statistics 2 ssl 51 Interpret SCS Statistics 0 AA e he 52 Interpret cluster wide Lock Statistics ee 54 Interpret cluster wide CI NI and Adapter Statistics cee nee 55 Interpret cluster wide Disk Statistics e nena 56 Interpret clu
552. th hard and soft during the reporting interval This includes system faults System Faults The number of page faults incurred in system space per second during the reporting interval The following are examples of system components that are pageable in system space XQP caches logical name tables process page tables global page table and some OpenVMS executive code for example RMS System Fault is a special designation for a page fault in addition to the types of faults described below Chapter 3 Evaluate Performance in Detail 79 Tabular Report Sections Pages Read The number of pages read per second to resolve page faults during the reporting interval Note that this may be from a page file an image file or a file backed global section Read I Os The number of page read I O operations per second during the reporting interval Pages Written The number of pages written per second to disk during the reporting interval including pages written to the swapping file to mapped image sections and to the paging file for modified page writing Not included are pages written to user files RMS Write I Os The number of page write I O operations per second during the reporting interval Free List The number of page faults per second resolved from the free list during the reporting interval Modify List The number of page faults per second resolved from the modified list during the reporting interval Dzero Fa
553. the Panel menu and click the Copy menu item The Copy Panel dialog is displayed as shown in the following screen Type the new panel name and click the OK button Delete panels Click a panel name to select a panel Pull down the Panel menu and click the Delete menu item The following message box is displayed as shown in the following screen f Do you really want to delete the Panel If you click OK that panel is deleted Chapter 10 Customize the DECwindows Motif Real time Display 343 Specify Actions on Panels Auto Startup Enable or disable a panel to startup automatically When you enable automatic startup for a panel it will be displayed and an attempt to connect will be made when you invoke the Real time Display software You can have several panels set for automatic startup To enable a panel to startup automatically 1 Select a panel from the Instrument Panel Directory in the Panel Manager by clicking on the panel name 2 Choose Enable from the Auto Startup menu The Instrument Panel Directory is updated The label auto startup is appended to the panel name To disable a panel from starting automatically 1 Select a panel from the Instrument Panel Directory in the Panel Manager by clicking on the panel name 2 Choose Disable from the Auto Startup menu The Instrument Panel Directory is updated The label auto startup appended to the panel name is removed Auto Connect Enable or disable
554. the data that must be displayed Performance Manager Factory Evidence The Performance Manager s factory rules use column headers that are created to improve readability If you need to reference the same data cell that is displayed as evidence on a factory rule conclusion reference the factory rules file PSPASEXAMPLES PSPASKB VPR to see what evidence data cells are provided for the desired rule These factory rules have special evidence processing M0010 M0050 M0055 M0060 M0421 R0140 R0150 R0160 R0170 R0210 R0230 R0240 R0245 R0250 R0270 R0280 L0040 L0050 L0060 L0070 L0080 L0090 The special processing allows non chronological sorting and the use of special display requirements These rules do not have evidence data cells listed in the factory rules file The evidence headers and conclusions for factory rules are stored in the file SYS SYSTEM PSPA MSG TXT See the Data Cell Types and Use section for a list and description of the data cells Chapter 5 Customize the Knowledge Base 155 Data Cell Types and Use Conclusion Text Element A conclusion is a block of text presented in the Analysis Report when a rule fires It describes the problem detected by the rule The format for conclusion text is shown in the following code CONCLUSION text string Conclusions can contain multiple lines of text An example of a multiline conclusion follows CONCLUSION Queues are forming on heavily used disks Longer delays will
555. the top five plus Other Users users that had the largest combined virtual address space by user name The metric graphed is VASIZE TYPE TOP VA WORKLOADS Top VA Space Workload Plots the top five plus Other workloads that had the largest combined virtual address space The metric graphed is VASIZE TYPE TOP WSSIZE IMAGES Top WS Size Images Plots the top five plus Other Images images that had the largest combined working set sizes by image name The metric graphed is WSSIZE TYPE TOP WSSIZE USERS Top WS Size Users Plots the top five plus Other Users users that had the largest combined working set sizes by user name The metric graphed is WSSIZE TYPE TOP WSSIZE WORKLOADS Top WS Size Workload Plots the top five plus Other workloads that had the largest combined working set sizes The metric graphed is WSSIZE ADVISE PERFORMANCE GRAPH The default graph of CPU UTILIZATION for today is displayed for all nodes ADVISE PERFORMANCE GRAPH TYPE TOP RESPONSE TIME VOLUME _ NOSTACK FORMAT REGIS CHARACTERISTICS LINE The previous command produces a graph of the top response times for the top 5 disks NOSTACK and LINE are used together to compare the response times on a graph without any occlusion ADVISE PERFORMANCE GRAPH COMPOSITE BEGINNING 9 JAN 1990 09 00 ENDING 10 JAN 1990 09 00 TYPE TOP IO VOLUME HISTORY monthly user 230 Performance Manager Administrator Guide ADVI
556. ther collection definitions can be limited to only those classes of data needed for special purpose analysis Class coverage All classes d Selected classes WB eve All processes a Modify process list mana B o data All devices o Modify device list The toggle buttons enable you to choose all or selected classes All classes Turn on this toggle button to collect all classes of data Selected classes Turn on this toggle button to choose which classes of data you wish to collect processes IO data or metrics Chapter 8 Use the DECwindows Motif Interface 301 How You Customize Processes Turn on this toggle button to collect Process data or to collect process data for specified processes IO data Turn on this toggle button to collect IO data or to collect IO data on specified devices Metrics The Metrics field lets you choose whether or not to collect a summary set of system metrics including such data items as memory utilization CPU and I O parameters All processes When you select the collection of process data the All Processes option is collected by default If you have modified the process list the menu is revised to reflect the type of coverage list include or exclude and the count of specified processes Clicking on this menu will display the correct list This menu provides a convenient way to view the coverage list Modify process list Click this button t
557. throughput rate through the adapter nexus port that is experiencing excessive throughput on the local node for the current interval or a zero It is zero if the Boolean data cell EXCESS THRUPUT ON ANY CHANNEL is zero Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP CHANNEL OVER THRESH TYPE Derived This contains a text string representing the type of I O adapter such as CI MASSBUS that is experiencing excessive throughput on the local node for the current interval or None It is None if the Boolean data cell EXCESS THRUPUT ON ANY CHANNEL is zero Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO TAP Appendix C Performance Manager Data Cells 421 Performance Manager Data Cells COMMUNICATION SCAN Derived Provides the count of communication subrecords for which the specified rule expression is true The expression will be evaluated for each communication subrecord Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains COM COMM_CONTROLLER_NAMECOM_A_CTLR_NAME This contains a string indicating the terminal controller of the current communication subrecord for the current interval Data Type STRING Domains COM COMM_OPERATION_RATECOM_F_OPCNT This contains a value representing the average operations rate for the current communications subrecord on the local node for the current interval Data Type NUMERIC Domains COM COMM_OPERATION_RATE_TALLY Derived
558. tics in the Performance Evaluation Report can be presented in a number of ways The previous example showed the default presentation of process statistics The following example shows the data presented by image and user This report was generated with the PROCESS_STATISTICS FOCUS TRADITIONAL PRIMARY_KEY IMAGE SECONDARY_KEY USERNAME qualifier The following statements are keyed to the columns in following example 1 Identifies the image name The Performance Manager displays process statistics for each image executed 2 Identifies all of the users who activated the image 3 The imagename with a summarization of its overall usage 38 Performance Manager Administrator Guide Performance Evaluation Report Performance YQUEM VAX 6000 440 Evaluation Tuesday 26 JAN 1997 09 00 to 10 00 The table below lists observed workload characteristics of all the interactive images that were run during the given interval Note that Diskio Bufio and Cputim are percentage contributions of the respective images to the total system load Working set size and working set faults are the average for the respective images In the case of 0 image activations the Uptime image and Cputim image actually report the cumulative Uptime and Cputim for the image Node Name YQUEM IMAGENAME dcl of Page Faults Avg of of Uptime Cputim activ per Actvtn Ws Direct Buffered of image image ations Soft Hard size 1 0 I O Cputim STEWART SYSTEM
559. tilization CPU utilization for this Percentage Disk IO Rate Disk Thruput Hard Fault Rate Soft Fault Rate process Disk I O for this image Disk throughput Hard page fault rate Soft page fault rate 332 Performance Manager Administrator Guide Operations per second Kilobytes per second Faults per second Faults per second Memory Panel Default Panel Descriptions Memory Allocation Panel Label Metric Units Next Panel Buf I O Rate Buffered I O rate Operations per second Direct IO Rate Direct I O rate Operations per second Label Metric Units Next Panel Free List Pages Number of pages on free Pages Memory Allocation list Total Memory Memory by user name Pages User Memory Pages by Users list Page File Page file utilization Percentage Page File Utilization Allocation Label Metric Units Next Panel Free List Fault Free list fault rate Faults per Rate second Top Memory Memory by process name Pages Processes list Paged Bytes Number of pages in paged Pages pool NonPaged Bytes Number of pages in Pages nonpaged pool Free Balance Number of free balance Count Slots set slots IRPs Used Number of IRP packets in Count use SRPs Used Number of SRP packets in Count use LRPs Used Number of LRP packets in Count use The fields marked with the are obsolete and are set to zero Chapter 9 Use the DECwindows Motif Real time Display 333 Default Panel Descripti
560. ting Customizing Displays the process definition of the logical name PSDCSDATABASE if it exists If not the job definition group definition or system definition is displayed If you want to analyze a Performance Manager database directory other than the system directory you can specify an alternate directory This directory might contain Performance Manager data files from another cluster or archived data whose classification definition is not applicable to your current scheme Performance Manager Status The status of the data collection process can be one of the following Running Stopped Down Waiting due to schedule Waiting for disk space No path to database device Unknown user lacks SYSLCK privilege For more information on Performance Manager status see the Performance Manager Administrator Guide Data Selected for Processing Displays the start and end time of the analysis period the processing options chosen and the nodes selected for analysis Until you select data the message No data selected is displayed and all display menu options are disabled This is a brief list of the results of an analysis of performance data Additional information can be gained by reviewing the Work in Progress box and the Data Selection box Chapter 8 Use the DECwindows Motif Interface 259 How You Control the DECwindows Interface Files locked by this session Displays one of the following files m Schedule file m Parameters
561. tion on how to do this see How to Implement Changes in this chapter However if the evidence presented does not look normal or if you don t know what normal looks like for those data items you need to investigate further One good way is to generate graphs of those or related data items for the time periods given in the evidence and then look for unusual spikes of activity If you find spikes occurring at the same time in different graphs then you might suspect that the data items graphed are somehow related to the underlying problem which might lead you to ideas of what might be changed to fix the problem 136 Performance Manager Administrator Guide Components of Rules Another path of investigation is to look for related rule firings A rule might not present the full picture by itself but when coupled with related rules could give you a clear indication of the source of the problem Related rules need not appear one after the other as Performance Manager processes rules in the following order CPU memory I O resource miscellaneous and after all nodes have been analyzed cluster wide You do not need to have a cluster for one of these rules to fire but the metrics used are not specific to an individual node so they are evaluated after all node specific rules Rules which require special evidence processing as indicated by the absence of evidence data items in the rules source file are presented first Finally if you obs
562. tions To set the panel options 1 Pull down the Options menu of the Instrument Panel 2 Choose the menu item you want 366 Performance Manager Administrator Guide How You Set the Panel Options The Options menu lets you perform the following actions Set the Panel Status Set panel status Specify panel background Specify panel title Specify panel node and instance type Remove panel menu To set the panel status 1 2 3 Pull down the Options menu Choose the Status Display menu item Choose either the Restore or Remove submenu items The panel status is either displayed or removed from the panel s lower border The Panel Status displays the following items Panel mode Connected indicated by a circle enclosing a vertical line Disconnected indicated by a circle enclosing a broken vertical line Last instrument update time if connected Interval in minutes and seconds Specify the Panel Background To specify the panel background Pull down the Options menu and choose the Panel Background menu item Performance Manager displays the Pattern Editor See the section Set Patterns and Colors see page 349 section for a discussion on how to use the Pattern Editor Chapter 10 Customize the DECwindows Motif Real time Display 367 How You Set the Panel Options Specify a Panel Title To specify a title 1 Pull down the Options menu and choose the Panel Title menu item Performance Ma
563. tions Disk Volume Plots the five plus Other disk volumes that have the highest split I O operations The metric graphed is SPLITIO Chapter 6 Performance Manager Commands 227 ADVISE PERFORMANCE GRAPH TYPE TOP SUPER PROCESSOR Top Supervisor Mode Processor Plots the five plus Other processors in terms of time spent in Supervisor mode as a percent of CPU time The metric graphed is P SUPER TYPE TOP TERMINAL INPUT IMAGES Top Image Terminal Input Plots the top five plus Other Images images with the highest character per second terminal input The metric graphed is TERM INPUT TYPE TOP TERMINAL INPUT USERS Top User Terminal Input Plots the top five plus Other Users users with the highest character per second terminal input The metric graphed is TERM INPUT TYPE TOP TERMINAL INPUT WORKLOADS Top Workload Terminal Input Plots the top five plus Other Workloads workloads with the highest character per second terminal input The metric graphed is TERM INPUT TYPE TOP TERMINAL THRUPUT IMAGES Top Image Terminal Thruput Plots the top five plus Other Images images with the highest character per second terminal thruput The metric graphed is TERM THRUPUT TYPE TOP TERMINAL THRUPUT USERS Top User Terminal Thruput Plots the top five plus Other Users users with the highest character per second terminal thruput The metric graphed is TERM THRUPUT TYPE TOP TERMINAL THRUPUT WORKLOAD
564. trics selected by image name Process metrics selected by workload name Disk device metrics selected by device name Disk device metrics selected by volume name Processor mode metrics by physical processor ID HSC metrics by HSC node name SCS metrics by SCS Cluster node name RULE metrics by rule id HSC Channel metrics by channel name File metrics by file name IOs by Username and Volumename IOs by Imagename and VolumenameEnter Choice 1 14 2 o O TO UI Bb WN H Enter Choice b 1 14 2 1 Select up to 6 User Names lt CR gt for list SANTELK HAFFMAN SARBID ORIPAS FRED CARL DIEGER 2 Select one process metric CR for list 3 GRAPH TYPE CUSTOM Process metrics The following process metrics can be requested on a custom graph BUFIO CPUTIME DIRIO DSKIO DSKTP FAULTS HARDFAULTS IMAGE ACTIVATIONS IO SIZE RESIDENCE RESPONSE TIME TAPIO TAPTP TERM INPUT TERM THRUPUT WSSIZE VASIZE Select one process metric CR for list CPUTIME 4 Enter an optional title for the CUSTOM graph 40 characters max Title PSPA CUSTOM GRAPH CPU time used by the team 5 PSPA gt 132 Performance Manager Administrator Guide Graph the Hot File Activity CPU time used by the team ie YQUEN FEE 97 121 13 43 Pee Rack Percent of CPU SSTTV A QTXOVIS Graph the Hot File Activity By default PA combines files with the same names regardless of differing directory or device locations To separate
565. ts Disk analysis PC sampling Real time file activity display Chapter 1 Introduction 13 Performance Manager m Performance Manager makes recommendations for improving system performance It does this by analyzing system data through the application of expert system technology identifying specific conditions causing performance degradation and presenting detailed evidence to support its conclusions It also provides real time displays of performance data using either DECwindows Motif or character cell displays The user can interactively view and investigate system performance problems and resource usage The following functions are included in this component Performance knowledge base and rules compiler Performance analysis and reports Real time displays of performance data Graphing Data export facility The common utilities that are shared by all components provide the capability for managing and interrogating the files in the database These include a workload parameter editor a schedule file editor a data archive and a data file dump utility Performance Manager To deal with system performance effectively you must understand the workload and the capabilities and limitations of system resources Generally any attempt to improve system performance requires specific performance goals stated in measurable terms Performance Manager provides the tools you need to analyze graph and present performance
566. u receive the following prompt Type CR to continue If a broadcast message disrupts the display of the ReGIS graph you can refresh the display with the following procedure You can either enter a C for color P for pattern or L for line to regenerate the graph currently on the screen with a changed characteristic The DCL qualifiers that obtain these characteristics are as follows FORMAT ReGIS CHARACTERISTIC COLOR FORMAT ReGIS CHARACTERISTIC LINE FORMAT ReGIS CHARACTERISTIC PATTERN With each method it is possible to redraw the current graph if a broadcast message disrupts the display Chapter 4 Generate Historical Graphs 117 Format Graphs and Pie Charts Output Formats The following figure PostScript Graph illustrates a PostScript formatted pattern graph The graph was generated with the following command ADVISE PERFORMANCE GRAPH BEGINNING 19 FEB 2006 13 00 ENDING 19 FEB 2006 17 00 _ TYPE CPU UTILIZATION NODE GALLO OUTPUT CHACPU PS _ FORMAT POSTSCRIPT CHARACTERISTIC PATTERN CPU UTILIZATION N10 Node GALLO Date 19 FEB 1997 13 00 17 00 48 Pt Each 5 0 100 LIN 60 Percent of CPU 40 SSIYIVA CHNOV IS Min 16 47 Max 100 00 Average 62 51 Batch E Detached pn Interactive REE Inisk MPeynch B B The following figure PostScript Formatted Line Graph illustrates a PostScript formatted line graph The graph was generated with the following command ADVISE PERF
567. ubrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG KB_RECEIVED_VIA_REQST_DATASSCS_F_KBYTREQD This contains the value representing the number of kilobytes transferred per second via request data s to the local node from the remote node for the current configuration record and interval Data Type NUMERIC Domains CFG KB RECVD VIA REQST DATAS TALLY Derived This contains the sum of the values representing the number of kilobytes transferred per second via request data s from the local node to the remote node for all the current configuration subrecords which were selected by the most recent CONFIGURATION SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains CFG KB SENT VIA SEND DATASSCS F KBYTSENT This contains the value representing the number of kilobytes transferred per second via send data s from the local node to the remote node for the current configuration record and interval Data Type NUMERIC Appendix C Performance Manager Data Cells 461 Performance Manager Data Cells Domains CFG KB SENT VIA SEND DATAS TALLY Derived This contains the sum of the values representing the number of kilobytes transferred per second via sent data s from the local node to the remote node for all the current configuration subrecords which were selected by the most r
568. ueuing time at the server and the service time It is measured from the time the I O request is issued until the time the controller completes the request Queue length The average number of disk I O requests waiting for service Space Used Yo The percentage of the total disk volume space that is allocated Reported only for file structured and mounted disks The following example shows the Server Statistics Section Server Statistics Work Avail Paging Swaping Queue 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 The server statistics reported are as follows Work Avail 9 o The percentage of time there were I O requests at the server queue Reported only for By Node statistics Chapter 3 Evaluate Performance in Detail 89 Tabular Report Sections Paging The percentage of the work available marked as Page I O Swping The percentage of work available marked as Swap I O Queue Length The average sum of the requests at the server queue Process Metrics Data Process metrics consist of m Standard process metrics with image name m Extended process metrics Standard Process Metrics The following example shows standard process metrics data Standard process metrics are collected by the Performance software from process data structures and include these statistics PID The process identification in hexadecimal Process Name The process name UIC The process user identification code Pri The process
569. ular report contents e 235 Task CPU e 390 Terminal ReGIS e 373 response time graph e 191 VT100 e 373 VT240 e 373 TERMINALS graph type e 191 Terminating Real time sessions e 345 THRESHOLD keyword e 245 Threshold value modification e 18 Throughput highest disk device e 191 disk volume e 191 files e 191 HSC channel e 191 HSC disk e 191 images with highest e 191 Index 587 users with highest e 191 workload with highest e 191 TITLE keyword e 191 245 TODAY keyword e 191 TOMORROW keyword e 191 Top BDT Wait Rate e 191 Blk Transfers Requested graph e 191 Transfers Sent graph e 191 Buffered 1 O Images graph e 191 1 O Users graph e 191 I O Workloads graph e 191 Busy Disk Devices graph e 191 Disk Volume graph e 191 Physical Processor graph e 191 Cluster Rule Occurrences graph e 191 Compat Mode Processor e 191 CPU Images graph e 191 Rule Occurrences graph e 191 Users graph e 191 Workloads graph e 191 Credit Wait Rate graph e 191 Datagrams Discarded graph e 191 Received graph e 191 Sent graph e 191 Direct I O Images e 191 Users graph e 191 Workloads graph e 191 Exec Mode Processor e 191 Faulting Images graph e 191 Users graph e 191 Workloads graph e 191 Freeblks Disk Device graph e 191 Disk Volume graph e 191 Hard Faulting graph e 191 Users graph e 191 Workloads graph e 191 HSC Channel 588 Performance Manager Administrator Guide I O graph e 191 Thruput e 191 HSC Disk IO graph e 191 Thruput graph e 191 HSC
570. uling sse a er nnn 114 Use Binary Graph Data aan nG a tea eye NADAAN NONA atre area vate e es 114 Components of Pie Charts 2 sls nn 115 Pie Chart Presentation of CPU Utilization s rs 116 Format Graphs and Pie Charts 222 22 tte rl 116 Refresh a ReGIS Graph with New Characteristics esee 117 Output FORBES eio NA TATAG 118 Data Resolution with X POINTS 222 20 le 123 Generate Custom Graphs 2 22 22 sls nn 126 Graph System Metrics sh re 127 Graph Process Metrics by User 2 re 130 Graph the Hot File Activity 2 sa 133 Chapter 5 Customize the Knowledge Base 135 The Knowledge Base 2 2 2222 222 sae n nl 135 Investigate Rule Firing sssr rris eec e 393 Gd a d P GGd d E qud e de ed vp pe 136 Components of Rules 0 cece AA NAA NG AN NG NA eh hh aaa aNG 137 RUES FE Consthucts ever hve ree cere eee NAUNANG TENTANG WAN ANNE Um 137 Rule Construct Elements sesers GG AN hh eh s 148 Data Cell Types and Use 2 sen 156 Boolean Data Cell 32 0320 24080046 4 ta ti istini aa a i s ip gp ee DE REVERIE TERE DERE aqu ds 158 N meric Data Cell e 159 String Data Cell 0 a 159 6 Performance Manager Administrator Guide Time Data Cell useseeeeeeeeeeee Rae 159 Scan Routine Data Cell 4 aaa eee eee erent ees eee yh EEEN 159 Tally Data Cell reer e a E E aa E E E EE Ea EE E EAE E ESE EEA E
571. ults The number of page faults per second resolved as demand zero pages during the reporting interval Gvalid Faults The number of page faults per second resolved as valid global pages already in memory during the reporting interval WritIn Prog WritIn Prog faults are the number of page faults per second resolved from pages currently being written to disk Hard Faults The percentage of Page Faults that required a read from disk This is read I Os page faults expressed as a percentage Soft Faults The percentage of page faults that were resolved from memory that is without reading from disk This is equal to 100 hard faults percent 80 Performance Manager Administrator Guide O Rates File 1 O Rates Tabular Report Sections This section describes the I O subsystem collected from OpenVMS performance data Direct I Os The number of direct I O operations performed per second during the reporting interval exclusive of page and swap I O This system wide statistic is also exclusive of I O to mapped image sections but includes RMS I O Buffrd I Os The number of buffered I O operations system wide performed per second during the reporting interval Lognam Trans The number of logical name translations system wide performed per second during the reporting interval Mailbx Reads The number of mailbox reads system wide performed per second during the reporting interval Mailbx Write The nu
572. umber of hot files to list per disk volume in the Hotfile Statistics section of the Performance Evaluation Report By default the maximum number of hot files is 20 250 Performance Manager Administrator Guide REPORT INCLUDE section Specifies which sections of the Performance Evaluation Report are to be included The negatable options are as follows m ALL STATISTICS m POOL STATISTICS m LOCK STATISTICS m TAPE STATISTICS m SCS STATISTICS m HOTFILE STATISTICS m CI NI AND ADAPTER STATISTICS m DISK STATISTICS m SUMMARY STATISTICS m PROCESS STATISTICS m MODE STATISTICS m RULE STATISTICS By default the Performance Manager includes all sections except SCS statistics Rule statistics are only available with archived data NODE NAME nodename Specifies the preparation of a report for only one of the selected nodes By default the Performance Manager prepares reports for all selected nodes cluster wide statistics are always included Chapter 7 Use Command Mode Commands 251 REPORT OUTPUT filespec Specifies an output file specification as a destination for the report output By default the output file destination is SYSSOUTPUT PROCESS_STATISTICS FOCUS TRADITIONAL SUMMARY GENERAL MEMORY_RELATED IO RELATED CPU RELATED PRIMARY KEY MODE USERNAME IMAGENAME UIC GROUP PROCESS NAME WORKLOAD NAME ACCOUNT NAME PID SECONDARY KEY MODE USERNAME IMAGENAME UIC GROU
573. ument s current metric name and class Hold MB1 down on the class name option menu to see all available choices Release MB1 on the desired class name A list of applicable metrics is displayed Click the metric name you want to display See Appendix C for a description of each metric Once you are satisfied with the metric selection for the field and you wish to specify another field in the instrument press Apply and then select another field by clicking with MB1 within the dashed outline defining the field 360 Performance Manager Administrator Guide How You Edit the Panel Instruments A sort option can be selected for all metric classes other than System To display only the data that matches a specific filtering criteria 1 Select the Filter toggle This makes available the Metric button 2 Click Metric to display the Filter Metric Selection dialog type oy Valeo S image Name Process Name Filter Equal a Account Process ID Metric Type STRING Mode First Compare Value Second Compare Value The Filter Metric Selection dialog lets you perform the following actions m Select a filter type m Select a filter metric name m Select the filter criteria m Specify filter values To enable filtering position the mouse cursor over the Type option menu and hold down MB1 The following options are displayed m No Filter m By Value m By Instance To enable filtering based on a metric value select th
574. ummarized and presented by the device key The device statistics are presented in descending order by the selected metric value To summarize the device statistics according to a different key m Press KP6 The available device keys areas follows m TOP Volume metric m TOP Disk metric m TOP Server metric To change the metric displayed m Press KP9 The available metrics are as follows m IO Rate m KB per second m Read Rate m Write Rate m Page Swap IO Rate Display Process Information Performance Manager provides a Process display when you press KP7 from the single node display when the Process Statistics section is selected or when you enter any of the following commands to the PSRT prompt m USERNAME name m IMAGENAME name m PIDn Chapter 11 Use the Character Cell Real time Display 381 Display Process Information The following illustration is an example of a user name Process display File Edit Commands Options Print Top User Cpu VQUEM Process Data at tine 09 0 HOF FHAN HOFFMAN WOTESSMA IN POR FE M S DECWIDOOKREADER 3 HOFFMAN EMACS F HOFFMAN PSPASMO 11f 761 HOFFMAN tMAC In the upper left corner is a line describing the current summarization and sort position of the Process information This text follows the format nth TOP key metric where you select the key by pressing KP6 either User Image Processname Accountname or PID and metric is selected by KP9 see the previous section for a list of a
575. uring the reporting time period On an SMP system all active processors are considered when computing this value Average percentage of each node s memory that was used during the reporting time period Average number of interactive jobs by node during the reporting period Average number of batch jobs by node during the reporting period Average number of network jobs by node during the reporting period Average number of detached jobs by node during the reporting period Command line used to generate the requested report Chapter 3 Evaluate Performance in Detail 63 Histograms Histograms Histograms provide a chronological view of the CPU memory disk and terminal I O use for each node as well as node status information Select histograms by specifying the HISTOGRAM option in the command line as follows ADVISE PERFORMANCE REPORT HISTOGRAMS The data in histograms shows how the system is being used during the specified time interval A shorter reporting period alters the scale of the histograms providing finer resolution This information helps you double check some of the conclusions reached by the Performance Manager including CPU and memory limitations Image Residence Histograms You can also plot the residence time for a specified interactive image The residence time is the time in seconds between image activation and image termination This information can help you track images that consume a fixed amou
576. urrent disk on the local node for the current interval Data Type NUMERIC Domains DSK 438 Performance Manager Administrator Guide Performance Manager Data Cells DISK MSCP PAGING IO TALLY Derived This contains the sum of the values representing the average MSCP served paging I O rate for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_MSCP_THRUPUT_RATEDEV_F_MSCPIO This contains a value representing the average number of bytes per second transferred to and from the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_MSCP_THRUPUT_TALLY Derived This contains the sum of the values representing the average MSCP served throughput rate for the current disk subrecords which were selected by the most recent DISK_SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK OVER QL THRESHOLD X Derived This contains an index pointing to the disk subrecord experiencing an excessive queue length on a disk on the local node for the current interval It is set up when the cell ANY DISK OVER QL THRESHOLD becomes true Data Type INDEX Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK OVER THRESHOLD X Derived This contains an index pointing to the disk subrecord experiencing an exce
577. vailable metrics To proceed to the nth 1 entry press KP1 or KP4 to go back To lock the display on a specific process enter PID nnnnnnnn for the desired PID Also in the upper left corner is the process identification section user name image name process name account name PID and process mode for the currently displayed process If any of these fields have an asterisk more than one process has been summarized for this screen and the given field had more than one value The following screen is an example of the Performance Manager Real time Process Display Single Process fie Edit Commands Options Print t Top PID Cou 29402316 2 7 100 YSTEM T YOUEM P R T PSPREIRT MOTIF TERA TIVE 382 Performance Manager Administrator Guide Display Process Information In the upper right corner is the name of the node and the current time of the data being viewed Next on the right side are three bar charts that show CPU utilization memory utilization and the processes working set as a percentage of the total system s memory and the number of disk I Os as a rate per second The mid section of the screen contains the process statistics for the selected process as in the previous illustration If more than one process matches the selection criteria such as a given user name the working set data and state information is replaced by a list of processes that match the criteria as in an earlier illustration Performance M
578. val Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains PRO PROCESS STATEPROA A STATE This contains a string representing the process s scheduling state at the end of the sampling interval This cell pertains to the current process subrecord on the local node for the current interval Data Type STRING Domains PRO Appendix C Performance Manager Data Cells 485 Performance Manager Data Cells PROCESS STATUSPRO B STATUS This contains a code representing the process s status where 0 means interactive 1 means batch and 2 means network This cell pertains to the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS TAPE IO RATEPRO F TAPE IO This contains a value representing the tape I O rate of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS TAPE THRUPUTPRO F TAPE THRUPUT This contains a value representing the tape throughput rate of the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS TERM INPUTPRO F TERM INPUT This contains a value representing the rate of terminal inputs for the current process subrecord on the local node for the current interval Data Type NUMERIC Domains PRO PROCESS TERM RESPONSE TIMEPRO F RESPONSE TIME This contains a value representing the average number of milliseconds b
579. value is an average weighted by the processes uptime m WS Extent This column provides the WSEXTENT value for the process For more than one process the value is an average weighted by the processes uptime View Process Statistics with an Emphasis on CPU Metrics The following report example illustrates the process statistics section with a focus on CPU related statistics The default primary and secondary keys are USERNAME and IMAGENAME This format of the report is obtained by using the qualifier PROCESS STATISTICS FOCUS CPU RELATED To view the report with a different orientation provide your choice of sort keys For example the following syntax presents the statistics by imagename with a breakdown of who is using those images PROCESS STATISTICS FOCUS CPU RELATED PRIMARY KEY IMAGENAME SECONDARY KEY USERNA ME 44 Performance Manager Administrator Guide Performance Evaluation Report Performance YQUEM VAX 6000 440 Evaluation Tuesday 26 JAN 1997 09 00 to 10 00 Primary Secondary Key CPU Cpu time Uptime Response Terminal Terminl COM SftFlt DskI O TapeI O Process User Image Utilizatn Min Min tim sec Inpt Sec Chrs sec Pri State Prct Sec Sec Sec Mwait lt dfs comacp gt DFS COM_ACP ARROYO dcl COPY DECPRESENT DIRECTORY MAIL BEYH dcl DECW QUOTE DQS CLIENT MAIL VUESMASTER User Command ADVISE PERF REPORT PERF NODE YQUEM BEG 26 JAN 1997 09 00 00 00 END 26 JAN 1997 10 00 00 090 INCLUDE PROC PROC
580. want to graph Click the Chart Wizard OR From the Worksheet Menu Bar click Insert Chart Select Chart Type Column with chart sub type from row three left sub type 3 D Column and then click Next Select Columns and then click Next Enter the title Processor Busy 562 Performance Manager Administrator Guide Time 13 00 Step 5 Create the Graph in Excel 7 Set the X axis as Time the Y axis as CPU and the Z axis as Yo Processor Total Busy and then click Next 13 08 9 To adjust the chart Tips for viewing your data in the Excel Chart In Step 4 of the wizard click As a new sheet for the chart location and then click Finish The chart appears on a new sheet in the current Excel file as shown in this example RX8620 Processor Busy RX8520 D RX8520 1 RX8520 2 RX8520 3 pop up menu RX8620 4 RX8620 5 0 8 0 4 0 2 o RX8620 9 RX8520 10 RX8620 11 CPU 06 Processor Total Busy E RX8620 0 m RX3520 1 O RX6620 2 o RXB520 3 RX0620 4 m Rx8620 5 B RXB520 5 O RXB520 7 m RX3620 8 m RXB520 9 o Rx8620 10 m RXB520 11 right click in the chart and choose an option from the Use the 3 D View to rotate the chart If the collection intervals are all on a single day you can choose to omit Column A from your graph selection Add the date to the worksheet title for the chart or to the Chart Title Appendix F How You Graph Seven or More CPUs 563 Glossary
581. while in the OpenVMS executive but not on the interrupt stack Exec Percentage of time the CPU spent executing in executive mode For example RMS is usually executed in executive mode Chapter 3 Evaluate Performance in Detail 77 Tabular Report Sections Lost CPU Super Percentage of time the CPU spent executing in supervisor mode For example DCL normally executes in supervisor mode User Percentage of time the CPU spent executing in user mode Compat Percentage of CPU time spent in PDP 11 compatibility mode Not all processors support compatibility mode hence in these cases this value is always zero System Computed as the sum of the interrupt busy wait kernel and executive CPU Busy percentages This metric represents the amount of CPU time the system uses to keep itself running and can be thought of as overhead Task Computed as the sum of the supervisor user and compatibility mode busy percentages This metric represents the amount of CPU time the system uses to perform work This section reports occurrences of the CPU s inability to execute because of some memory limitation Page Wait Percentage of time that the CPU was idle and at least one disk device had paging I O in progress In a multiprocessor system all CPUs must be idle Swap Wait Percentage of time that the CPU was idle and at least one disk device had swapping I O in progress This includes both swapping and modified page writing I
582. will permit process growth beyond the working set quota WSQUOTA for the process for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN CACHE STATEPAR F CACHE STATE The value of the SYSGEN parameter VCC FLAGS in combination with the state of the cluster and OpenVMS version allow this data cell to reflect one of the following states for the virtual IO Cache 1 indicates the cache is enabled 2 a heterogeneous cluster disables the cache 4 disabled cache 8 in determinant cache state cannot decode data structures 16 XFC is operating in FULL mode 32 XFC is operating in REDUCED mode Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN DEADLOCK WAITPAR F DEADLOCK WAIT The value of the SYSGEN parameter DEADLOCK WAIT which defines the number of seconds that a lock request must wait before the system initiates a deadlock search on behalf of that lock for the local node for the current interval record in LOCAL domain and for the local node for the last interval record in SUMMARY domain Data Type NUMERIC Domains COM CFG CPU DSK FIL LOC PRO SUM TAP SYSGEN DEFPRIPAR F DEFPRI The value of the SYSGEN parameter DEFPRI which is the default priority for job initiations for the current interval for the local node Data Type NUMERIC Appendix
583. wing sections list the graph types and their descriptions Included are keywords used with the TYPE qualifier TYPE CUSTOM You must specify the items for the Performance Manager to graph The metrics and selection objects are described below ADVISE PERFORMANCE GRAPH TYPE CUSTOM SYSTEM METRICS system metrics USER METRICS process metrics SELECTION usernames IMAGE METRICS process metrics SELECTION imagenames WORKLOAD METRICS process metrics SELECTION workloadnames DEVICE METRICS disk metrics SELECTION devicenames VOLUME METRICS disk metrics SELECTION volumenames CPU METRICS cpu modes SELECTION Phy cpu ids HSC METRICS hsc metrics SELECTION HSC nodenames SCS METRICS scs metrics SELECTION SCS nodenames RULE METRICS rule metrics SELECTION Rule ids CHANNEL METRICS channel metrics SELECTION channel specs FILE METRICS file metrics SELECTION file names DISK USER METRICS disk user metrics SELECTION username volumename DISK IMAGE METRICS disk image metrics SELECTION imagename volumename NO STACK Y AXIS MAXIMUM n THRESHOLD m TITLE string Chapter 6 Performance Manager Commands 203 ADVISE PERFORMANCE GRAPH Where metric class The metrics are grouped together by metric class and described in the next table Selection string Specify up to six strings or only one if you specify multiple metrics The strings are used to match against
584. wledge Base Changing a Rule Value An example of what you would include in your MYRULES VPR to effect this change follows DISABLE 10180 Literal TD HIGH ERASE IO 0 1 EndLiteral Rule UI180 Split io rate lt td high split io Window turn rate lt td high turn rate Erase QIO rate ge td high erase io Occurrences td io Evidence Erase QIO rate Time Conclusion Security erasures as measured by Erase I 0s have exceeded threshold These I Os may be generated on a per file basis by the use of DELETE ERASE or PURGE ERASE They may also be generated on a per volume basis through the use of SET VOLUME ERASE ON DELETE or by NOT turning off the HIGHWATER MARKING attribute which OpenVMS enables by default for all volumes Brief conclusion Security erasures detected disable if not necessary EndRule In the previous example the conditions and evidence were copied from the factory rules source file while the conclusion and brief conclusion were copied from the message text file Build an Auxiliary Knowledge Base You can supplement and disable Performance Manager rules To modify the effect of the default Performance Manager knowledge base create an auxiliary knowledge base with your own rule definitions Edit your rules using a standard text editor 172 Performance Manager Administrator Guide Use an Auxiliary Knowledge Base for Reporting and Archiving Build your auxiliary knowledge base by entering
585. wn of this value from contributing nodes when applicable 6 Name of the node that uses the various tape volumes For rows in which no source node name appears the data refers to the cluster wide activity on the tape volume 7 Name of the node that services I O requests 8 Percentage of time that I O requests are outstanding to the volume for each node utilizing the tape volume 9 Percentage of total I O activity devoted to read operations This value for each tape is followed by the percentage of read operations to the tape from contributing nodes when applicable 10 Percentage of total I O activity that was split I O operations This value for each tape is followed by the percentage of read operations to the tape from contributing nodes when applicable Chapter 3 Evaluate Performance in Detail 59 Performance Evaluation Report 11 Tape volume type In this example all of the tapes are TASOEs 12 Count of the Performance Manager records containing data for a tape volume during the interval The following example shows the cluster wide tape statistics section of the Performance Evaluation report Performance CLUSTER Page 2 Evaluation PA Vx x Wednesday 14 JAN 2006 09 00 to 19 20 The following tables gives the summary of all tape activity as seen by the indicated node Percent Metrics during active records Cluster Tape of records Device Node controller Unit with activity I 0s sec SUPPLY DEMAND NODEO1
586. wo different data types available for the exported file ASCII and BINARY The default type is ASCII which provides data in CSV comma separated variable format with text items in quotes For examples commands and the resulting output data see Appendix E To use either format in subsequent processing of the data two files have been placed in the PSPASEXAMPLES directory area PSPASDUMP ASCII DTR and PSPASDUMP BINARY DTR To specify a format supply a value to the qualifier of ASCII or BINARY The default is BINARY ADVISE PERFORM EXPORT TYPE BINARY CLASS NODEFAULT MEMORY NODE MYNODE BEGIN 9 00 _ END 10 00 0UTPUT MYNODE MEMORY BIN The previous command creates an export file containing memory statistics for node MYNODE for an hour s worth of data SADVISE PERFORM EXPORT TYPE BINARY _ CLASS NODEFAULT MEMORY NODE MYNODE BEGIN 20 Jan 1997 _ END 21 Jan 1997 INTERVAL 1800 0UTPUT MYNODE MEMORY 20JAN The previous command creates a file with a day s worth of data summarized into 30 minute records DTR DTR gt DEFINE DICTIONARY CDD TOP DECPS BIN DTR5SET DICTIONARY CDD TOP DECPS ASC DTR5 PSPASEXAMPLES PSPASDUMP BINARY DTR DTR gt DEFINE DOMAIN MEMORY BIN USING BINARY RECORD ON CONSMYNODE MEMORY 20JAN DTR gt READY MEMORY BIN DTR gt FIND MEMORY BIN WITH DATA TYPE MEMO 48 records found DTR gt LIST ALL HEADER AND MEMO The previous command shows you how you can examine the binary
587. xperienced when longer queues form Suggested Remedies 1 If the disk is fragmented use the command ADVISE COLLECT REPORT DISK SPACE devicename in order to check on this backup and restore the disk using the BACKUP utility without the PHYSICAL qualifier 2 If possible assign all new work during the noted times to other disk volumes 3 If possible attempt to lower the future usage of the noted disk volumes Volume name s Number of Samples PROBLEM DISK CONDITIONS 1 DISK QUEUE AT SERVER GE 1 00 OR MAXIMUM DISK QUEUE GE 1 00 AND DISK IS SERVED EQ 0 00 2 CW DISK IO RATE GE DISK IO RATE THRESHOLD OR CW DISK THRUPUT RATE GE DISK THRUPUT RATE THRESHOLD 3 OCCURRENCES GE 1 EVIDENCE Disk Volume Srv Node Time of Occ Shdw Rbld N Avg I 0 Avg IO Sz Src Node sec Que Pages Hottest File PROBLEM DISK MARCUS 5 SUPPLY E E PSCP010 B DEMAND 16 0 PSALL BIGFILE PROBLEM DISK MARCUS 12 00 SUPPLY d F 3 PSDC PSDC010 B DEMAND y s E PSDC PSDCO10 B PROBLEM DISK MARCUS SUPPLY E PSDC PSDC010 B DEMAND 7 s PSPA PSPA010 B PROBLEM DISK MARCUS SUPPLY PSDC PSDC010 B 21 DEMAND E i 2 SMITH WORK PSCP PSCP MAIN EXE 67 The following statements are keyed to the previous Cluster Rule analysis report example 1 Name ofthe disk experiencing heavy use 2 Name of the node that actually services requests 3 Beginning time of the interval in which the condition occurred 4 An ast
588. xtent cache for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_FID_CACHE_SIZEDEV_F_FIDSIZE This contains a value representing the number of entries in the file ID cache for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK_FREE_PAGESDEV_F_FREE This contains a value representing the average number of free pages on the current disk on the local node for the current interval Data Type NUMERIC Appendix C Performance Manager Data Cells 435 Performance Manager Data Cells Domains DSK DISK FREE PAGES TALLY Derived This contains the sum of the values representing the number of free pages for the current disk subrecords which were selected by the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK HAS A PAGING FILE Derived This contains a Boolean value representing true 1 0 if there is a paging file installed on the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK HAS A SWAPPING FILE Derived This contains Boolean a value representing true 1 0 if there is a swapping file installed on the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DISK HEADER CACHE SIZEDEV F HDRSIZE This contains a value representing the number of entries in the file header cach
589. y of the volume names or device names specified that record is selected When using one or more of the disk filters the following PA reports include only information on the selected disks m Disk Statistics section of the Performance Evaluation Report m Disk Metrics section of the Tabular report m Hot file report only files located on the specified disk are selected m Top or Custom Volumename and Devicename graphs m Top or Custom User Volume Image Volume and Workload Volume graphs m Top or Custom Hot file graphs only files located on the specified disk are selected If you specify both a process filter and a disk filter the hot file report section and hot file graphs will select only hot files that are both located on the specified disk volume and used by the specified process The same will be true for the User Volume Image Volume and Workload Volume graphs The following entries allow you to select specific processes or disks for the reports and graphs Usernames Specify a list of strings separated by commas spaces or tabs to generate reports and graphs for all process records with the username matching any of the specified strings Imagenames Specify a list of strings separated by commas spaces or tabs to generate reports and graphs for all process records with the imagename matching any of the specified strings Do not specify any trailing EXE nor the file version device or directory Processnames
590. y the most recent DISK SCAN routine operation Data Type TALLY Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_SCAN Derived Provides the count of disk subrecords for which the specified rule condition is true The condition will be evaluated for each disk subrecord Data Type SCAN Domains COM CFG CPU DSK FIL LOC PRO TAP Target Domains DSK DISK_SERVER_HWNAMEDEV_A_HWNAME This contains a string indicating the hardware name of the Integrity server or Alpha node which serves the current disk to the local node for the current interval If the server is an HSC this field is blank Data Type STRING Domains DSK DISK_SERVER_HWTYPEDEV_A_HWTYPE This contains a string indicating the hardware type of the cluster node which serves the current disk s data to the local node for the current interval e g HS50 ALPHA IA64 Data Type STRING Domains DSK DISK_SERVER_NODENAMEDEV_A_NODENAME This contains a string indicating the cluster node name of the node which serves the current disk s data to the local node for the current interval Data Type STRING Domains DSK 442 Performance Manager Administrator Guide Performance Manager Data Cells DISK SERVICE TIMEDEV F SERVICE This contains a value representing the average number of milliseconds between the I O events START IO and END IO for all I Os for the current disk on the local node for the current interval Data Type NUMERIC Domains DSK DIS
591. y the title to be the metric name overriding any supplied string 356 Performance Manager Administrator Guide Create Instruments How You Edit the Panel Instruments Title Font The Font Selection dialog lets you specify the following font characteristics Family Size Weight Slant Font options not available on your server will be disabled Metric Type Specifies a list of instances of one metric or a list of metrics For example to show the CPU Utilization for multiple processes on the system use the List of Instances option To show a CPU mode Interrupt Kernel Supervisor and so forth for the system choose the List of Metrics option since each metric has only one value associated with it Number of List Entries Specifies the number of meters to be displayed Label Fonts The Font Selection dialog lets you specify label font characteristics Data labels width in characters Specifies the space available for the data labels based on the characteristics of the font you have selected Metric fields width in characters Specifies the space available for the meter based on the characteristics of the font you have selected Units label width in characters Specifies the space available for the units label based on the characteristics of the font you have selected Set the toggle button to include a units label for the meters Specify the space for the units label in characters To create an instrume
592. yword as the unique file name For example ADVISE PERFORMANCE GRAPH NODE SYSDEV END 1 TYPE MEM CPU U CPU MODE OUTPUT REG PSPA I CREAGRAPHOUT PSPA Graph created file MUMMS DKA300 CORREY SYSDEV CPU UTILIZATION REG 1 PSPA I CREAGRAPHOUT PSPA Graph created file MUMMS DKA300 CORREY SYSDEV MEMORY UTILIZATION REG 1 PSPA I CREAGRAPHOUT PSPA Graph created file MUMMS DKA300 CORREY SYSDEV CPU MODES REG 1 SELECT GREATER_THAN percent LESS THAN percent NOSELECT Use SELECT in conjunction with the optional threshold values which may be specified on a per graph type basis If this qualifier is present before a graph is produced a check is made to see if the values to be graphed fall within the threshold values for the indicated percentage of points If so then the graph or pie chart is produced If not no graph is produced For details on THRESHOLD see the TYPE qualifier Keyword Meaning GREATER THAN percent At least percent of the graph points plotted must be greater than or equal to the threshold value specified with the TYPE qualifier LESS THAN percent At least percent of the graph points plotted must be less than or equal to the threshold value specified with the TYPE qualifier These keywords accept a single value representing the percentage of the points plotted that must meet the threshold criteria before the graph is produced Each graph point value is determined by the sum
593. yzed Data Daily Data CPD aJ Gassitie d Ey None Period to be Analyzed acl JE D elle SHE JHE start Time End Time in a Settime Default today o L schedule Sunday 0 24 Monday 0 24 Tuesday 6 24 Wednesi Eo a 5 L Calendar wodity Filter pr dity aa Processing Options WE Analyst Performance W tabular Final O Report Options W craphs l py tmagenames L By usr Img wrkd of Volume W Cy Usernames L y Selected Node CJ By Hot Filename s L Additional Options W ey disks channels CPUs HSCs workloads SCS nodes and rules The time period and the processing options selected affect the amount of time needed to complete the analysis and the quantity of memory required generally restricted by your pagefile quota PGFLQUO and the system parameter VIRTUALPAGECNT See the Appendix Estimate Virtual Memory Needs see page 535 for information on estimating virtual memory needs and selecting data Chapter 8 Use the DECwindows Motif Interface 267 How You Select Data for Analysis Including each graphing sub option expands the memory and CPU requirements Selecting archived data in place of daily data reduces memory and processing time but also reduces flexibility The By Selected Node menu item associated with graphing greatly increases memory requirements as does the number of x points The Performance Manager Data Selection dialog box lets

Performance Manager

Contents

Download Pdf Manuals

Related Search

Related Contents