The exploratory system control model multi
Contents
1. by Gateway Node 3 Gateway Node 3 sends the information part of the packet across the 1 3 link Gateway 1 in Loop 3 formats a packet having Loop Address 31 The packet bypasses Node 2 and Node 31 reads the packet An ACK packet is sent out on the loop using LID 10 and the packet is linked to the input queue for deliverance to Host A If Node 11 had not received an ACK message after a specified number of retransmissions it would utilize alter nate routing It would do this by marking the packet indi cating that alternate routing was used and changing the loop read address FAD from 3 to 2 Gateway Node 2 in Loop 1 would read the packet and send it across the 1 2 link Gateway Node 1 in Loop 2 would use an FAD of 3 The packet would bypass Nodes 21 and 22 and be read by gate way Node 3 The packet would be sent across the 2 3 link and gateway Node 2 in Loop 3 would use an FAD of 31 The acknowledgment message would be sent via the alter nate route Node 11 would also report to one or more network control processors who could remove the 1 3 link from service for repair This wouid invoive sending special broadcast contro packets to Loops 1 and 3 so that Link 1 3 would not be used The above method of indirect addressing can be used for resource allocation such that processes could be moved around the network so that spare or less utilized processors can be utilized For example let us say that Host E is brought down for service and t2. detected by the nodes and reported to a System Control Monitor node Al ternate routing is automatically performed when a node in another loop fails to ACK a packet after a specified number of retries Failure to ACK a packet is reported to the Mon itor node Queue overflows are also reported to the Monitor a queue overflow results from the failure of an external device to respond to the node Security A demonstration of the use of a Security Monitor node is performed using Loop 4 A CRT B776 conversation is mon itored by a PDP 11 70 to detect an invalid password The node connected to the PDP 11 70 is commanded such that it does a non destructive read on packets addressed to the B776 Thus the data sent by the CRT is read by both the B776 and the PDP 11 70 The PDP 11 70 monitors the CRT B776 conversation When a bad password is detected the Security node sends control packets to the nodes directiy upstream and downstream from the CRT node to perform loop around resulting in the CRT node being removed from the network Network architecture The ESM can be used to study System Control network architectures Since the logical connectivity of the network is maintained by the modifiable nodal tables the system can be used to model other network architectures Network From the collection of the Computer History Museum www computerhistory org 940 National Computer Conference 1979 control problems such as automatic channel rec
3. The exploratory system control model multi loop network by DANIEL J PAULISH Burroughs Corporation Paoli Pennsylvania PURPOSE The Exploratory Systems Control Model ESM Multiloop Network consists of the original three loop ESM network delivered in 1977 and the fourth Exploratory Systems Con trol Model Development ESMD loop delivered in 1978 The ESM also includes a fifth loop supplied under the Mod ular System Control Development Model MSCDM project The ESM provides a flexible tool for simulating and com paring a wide range of system control architectures and their related procedures and protocols The ESM has been de signed to model the class of the system control architectures that have the characteristics of decentralized operation modularity easy modification and upgrade capability high reliability high survivability and fail soft operation BACKGROUND The Defense Communications System DCS is a global multiple user system composed of leased and government owned transmission media relay stations and switching cen ters deployed in support of the National Command Author ities and the services including command and control in telligence and early warning as well as administrative and logistical communications In order to increase the reliability and availability of these DCS services it is essential to improve the responsiveness and robustness of the System Control SYSCON process as much as possible This r
4. de directions for further dialog as well as replies to previous messages The language is de signed to be modular so that it can be easily updated and HOST D enaPocs POP 11 70 CRT le Taper Ee p Ze SOLc HOST C B776 TCCF AUTODIN II HOST E POP tiv 63 Figure 3 ESM multi loop network From the collection of the Computer History Museum www computerhistory org enhanced In addition CRTs can communicate directly with the operating system of the particular processor as if it were a local terminal Messages are sent in the form of packets of length not greater than 256 bytes As each packet is sent the sending node holds it for acknowledgment ACK from the receiving node When an ACK is received by the sending node it frees the packet space If a non acknowledgment NAK is received the message is resent or sent by an alternate route Absence of an ACK or NAK after a timeout period is considered to be a NAK After a suitable number of resends without an ACK the message may be reported not sent Nodes automatically provide input and output queueing for the external device Sufficient extra memory space is provided in each node to permit receipt of system control commands from the loop and to act on these commands This is done to prevent a deadly embrace condition within the node If the input queue from the loop to the external device is full new input messages are rejected Room always e
5. del is to provide DCEC with the necessary integrated means to evaluate through hybrid simulation a variety of candidate SYSCON subsystem the architecture s thereby identified as being suitable for im plementation The technical and performance information obtained from the unified hybrid simulation model will ul timately be used in the preparation of performance specifi cations for the future DCS SYSCON subsystem LOOP OR RING COMMUNICATIONS SYSTEMS General operation A communications loop is a closed ring connected set of nodes providing data flow unidirectionally from one node to the next Each link between nodes is a single twisted pair of wires carrying a serial data stream in a self clocking code Full connectivity is achieved by associating a destination address with each packet of data carried on the loop A node to whom a packet of data is not addressed acts simply as a delayed repeater having no effect on that data other than introducing some delay The concept of a data ex change loop has been described extensively in the literature of computer communications by Reames Jafari et al Loops may be distributed such that each node contains its own power supply and cabinet and is located near the equip ment it interfaces or locally such that all nodes are connected within a single cabinet with cable connections to interfaced equipment A functional block diagram of a communication loop node is given in Figure 1 The L
6. dinating restoring and reporting functions necessary for effective techni cal supervision and control over trunks and circuits traversing or terminating in a facility ACKNOWLEDGMENT I wish to acknowledge that this work was supported in part by the Defense Communications Agency under Con tracts DCA100 75 C 0054 and DCA100 76 0081 I also wish to acknowledge the contributions of personnel associated with the project at the System Control Group at the Defense Communications Engineering Center and the Advanced De velopment Organization of the Burroughs Corporation in particular the guidance given by D Schutzer O Halpeny J Lynch and A Meyerhoff REFERENCES Reames C C System Design for the Distributed Loop Computer Net work Ph D Dissertation The Ohio State Univ Columbus Ohio March 1976 2 Jafari H A New Loop Structure for Distributed Microcomputing Sys tems Ph D Dissertation Oregon State Univ December 1977 3 Meyerhoff A D Paulish C Tomlinson and T Woodward Loop Com munications Systems Burroughs Corp Tech Rept TR77 3 Paoli PA November 1977 4 Burroughs Corp ESM User Manual Doc 66143 1 Paoli PA March 1977 5 Burroughs Corp ESMD User Manual Doc 66146 Paoli PA March 1978 6 Burroughs Corp Final Report for the ESMD Doc 64297 Paoli PA January 1978 7 Fluk M Technical Overview of the System Control Improvemen
7. e quirement demands a DCS SYSCON subsystem possessing such design features as modularity and fail soft operation Modularity implies a subsystem that is capable of being upgraded modified and reconfigured easily and fail soft implies a subsystem that tolerates partial failures yet is relatively immune to total collapse To afford these capa bilities the future DCS SYSCON subsystem is expected to consist of many semi autonomous mutually supportive _ geographically dispersed control centers During FY 75 Burroughs Corporation began development of an Exploratory System Control Medel ESM which cap italized upon the inherent flexibility of multiple intercon nected data transmission rings and microprocessor based host ring interface nodes to provide an initial capability for experimental validation of various candidate SYSCON sub system architectures characterized by distributed control 935 and graceful degradation under stress This capability to model apparently dissimilar architectures is a consequence of the universal physical connectivity provided by the ring structure coupled with flexible protocols that permit defi nition of different logical connectivities through selective routing of transmitted data In the broader context of the DCS SYSCON Program the longer term joint purpose of this effort and the separate but related Modular System Control Architecture Study and Feasibility Development Mo
8. hus Process 21 is to be moved to another processor Let us say that it is determined pos sibly by some bid quotation scheme that Host D of Loop 2 is to handle Process 21 In order to move the process control packets would be broadcast in each loop From the collection of the Computer History Museum www computerhistory org 938 National Computer Conference 1979 ESM ESMD IMPLEMENTATION System elements and connectivity The ESM is a communications system used to intercon nect devices e g terminals host processors data com munications lines so that each device can interface with any other device for information transfer To accomplish this each ring is supplied with nodes that act as interfaces from ring to device and from ring to ring The ring to ring nodes are called gateway nodes Each node is the same physically as any other node except for a small amount of special separable hardware for each type of node The major difference between nodes is in the software of the nodes The nodes provide all the necessary communications func tions of queueing parity checking ACKing NAKing re transmitting alternate routing etc The hosts and terminals need only supply the data processing functions and need not be concerned with the communications functions The gateway node interchanges are via cables in the ESM configuration but in principle can be via any communica tions medium such as telephone microwave relay
9. icate with the User Language on any processor The User Language consists of four major modes of op eration The first mode CRT to CRT provides users with the capability to send messages to each other This simulates communication between System Controllers at different The Exploratory System Control Model 939 sites who must talk to each other in order to isolate certain faults Mode 2 System Inquiry allows the user to examine the nodal configuration tables the tables are monitored on disk on a host computer Mode 3 System Control allows the user to modify the nodal configuration tables this feature provides the capability to model different network architec tures Mode 4 implements a distributed data base on the PDP 11 40s The TOTAL Data Base Management System is used to distribute records of files on the two processors The data base appears to the user to reside completely on one machine File transfer A file transfer utility has been written to transfer files between host processors The program allows peripheral sharing by providing a capability to send files to another machine s disk printer or tape Files can be obtained from another machine s disk In addition terminals can be AT TACHed to the various host processors in the system Fail soft operation The system is designed to tolerate partial failures LIU failures result in automatic loop back performed to remove the failure from the network The failure is
10. of variable size to be transmitted and allows multiple transmissions Loop clogging is possible in both cases however and special means must be employed to declog the loops under cer tain error conditions Loop clogging is the deadlock situation when packets cannot be written onto the loop until previ ously written packets are removed The Jafari loop is a double loop one used for control and the other for data The data loop is segmented such that a switched point to point circuit is set up when requests for communication are issued on the control loop The ESM uses a Newhall protocol at a loop frequency of one mega baud Simulation studies and queueing analyses indicate that this loop can support a throughput in excess of 750K baud without undue delay The Pierce DLCN and Jafari throughputs can be even higher due to simultaneous conversations Suppose the average node writes 15 packets of 2000 bits each per second for a total of 30 000 bits second At that rate a Newhall loop can support 25 nodes The worst case time that a node will have to wait for a write token is given by Eq 1 where M is the number of nodes 25 P is the packet size 2000 bits and C is the loop frequency of IM bits sec Thus Twr is 50 msec The average wait for a write token is given by vD where p is the loop utilization 0 75 for our example thus T 19 msec Multiple loops addressing schemes The ESM system has proved the capabili
11. on capability between devices in the heterogeneous system When a module fails the loop will recognize the failure and cut the failed module out of the system by forced loop back from the module s nearest neighbors The module may then be replaced and the loop will return to normal opera tion In the meantime the other modules will still be in operation so that degradation will be graceful in that only the operation of the failed module will have been lost Loop throughput capability Loop throughput total number of message units that can be sent over the loop per unit time without undue message delay to the receiving modules is a function of line speed loop discipline and the definition of undue message delay Various loop disciplines have been developed and com pared The Newhall loop which uses a special control packet called a Write Token can transmit only one message on the loop at a time and has the lowest throughput but has the advantage of simplicity and cannot be clogged by misdi rected messages It also shows some advantage in ease of detection and deletion of certain types of faults The Pierce loop which uses fixed size slots in which data packets can be placed can transmit multiple messages but the small fixed packet size causes greater overhead than in the DLCN loop The DLCN loop uses queues within each LIU that can expand or contract to hold upstream messages in temporary storage This allows packets
12. onfiguration can be studied Response time Since each loop frequency is independently modifiable response time studies can be done on the system The loop rates are modifiable via switches on the clock generator cards In addition for Loops 4 and 5 an external clock generator can be connected to drive the loop Software development The ESM can be used as a general software development facility Since each loop connected terminal can ATTACH to any processor software can be written on a variety of machines with different operating systems and different lan guage compilers In addition since files may be transferred between machines via the network duplicate copies of files can be kept on different machines Processors without cer tain resources e g line printers can utilize the resources of other machines via the network APPLICATION TO SYSTEM CONTROL The ESM Multiloop Network will be used as a DCS Sys tem Control Simulation Facility DCS System Control must accomplish the following functions e Network control Transmission and switched network configuration control which includes network and ex tension supervision reconstitution restoral and satel lite configuration control e Traffic control Control of traffic routing and tr fic flow Performance assessment of the DCS and status moni toring of the DCS resources Technical control Includes quality assurance and monitoring patching testing coor
13. oop Interface Unit LIU is responsible for reading data addressed to the node and writ ing data on the loop The Control and Interface Processor CIP is a microcomputer that provides a data communica tions interface to the external device The memory is used From the collection of the Computer History Museum www computerhistory org 936 National Computer Conference 1979 EXTERNAL INTERFACE PRIMARY LOOP SECONDARY LOOP Figure 1 Communication loop node for program storage routing tables and intransit queue stor age The external interface provides a hardware connection to the external equipment to be connected to the loop Modularity adaptability features The basic Burroughs loop node is a module made up of the LIU microprocessor CIP memory and external inter face The nodes are identical except for the external inter face and the external device interface software used to han dle the protocol between the microprocessor and the external device In the ESM external devices include pro cessors PDP11 40 PDP11 70 B776 terminals TD802 TD832 gateways between the multiple rings and data communication interfaces SDLC AUTODIN II TCCF The nodal external device interface software provides code conversion flow control intransit queueing logical attach ment capability and emulation of various communications protocols for the devices The interfacing capability of the nodal modules provides communicati
14. optical transmission or Satellite relay The terms loop and ring are interchangeable Each loop is housed in a separate cabinet in this implementation but this is not a necessity A loop could as easily extend throughout a building or facility CRT a g o o 9 m Bza i ait p o At ESE The ESM Multiloop Network is illustrated in Figure 3 Loops 1 2 and 3 were delivered in 1977 as part of the ESM Contract Loop 4 was delivered as part of the ESMD Con tract in 1978 Loop 5 was delivered as part of the MSCDM Contract in 1979 Features of the ESM The ESM is designed to be transparent to the user Re gardless of the CRT used and the host on which a particular activity takes place the activity will take place for the CRT that calls for it When a message is transmitted from a CRT suitable control bytes are added to the message by the CRT node and directed to the node of a nearby host When the host receives the message it will either handle the message completely if it can or it will pass it on to another host via the ESM for cooperative handling of the message This is done under program control using the content of the CRT message and the added control bytes The CRT will then receive a response from one of the hosts A CRT can AT TACH itself to any node in the network via user com mand Responses will generally be part of the user language which is designed to provi
15. t Pro gram DCEC Tech Rept TR4 78 Reston VA May 1978 From the collection of the Computer History Museum www computerhistory org
16. ty for providing multiple loop systems and has acted as a vehicle for testing multiple loop addressing schemes Figure 2 exemplifies a multiple loop system Three loops are shown connected via gateway nodes Gateway 2 of Loop 1 connects to Gateway 1 of Loop 2 via a hard wire connection independent of the loops Similarly Loop 1 connects to Loop 3 and Loop 3 connects to Loop 2 via gateways Each loop is independent of the other loops The small boxes are nodes and the numbers within the boxes represent the functional address of the node The functional address FAD is the local address unique within the loop In addition each node has a logical identifier LID unique within the system From the collection of the Computer History Museum www computerhistory org LOOP 1 2 1 G2 Z2 The Exploratory System Controi Modei 937 C 14 A a 2 jam a o 3 few tatoo s aw a 10 aT ALe gt ay lela Te Toya Figure 2 Indirect method of addressing An example of how alternate routing is implemented with a multi loop architecture using indirect addressing is given in Figure 2 Let us assume that Host Processor A on Loop 1 wishes to send a message to System Process 21 Host A sends a packet to its CIP with 21 as the destination LID and 10 as its source LID The CIP formats a packet using an FAD or loop address equal to 3 The packet is sent out onto the loop bypasses Nodes 12 and 2 and is read
17. xists for the receipt of ACKs and NAKs and other control messages These are acted upon with dispatch so that they do not reside in data memory for a long period If the output queue is full the external device is prevented from sending to the node The loop protocols are designed to be non blocking and self polling Each node in the loop has its turn to write onto the loop and if any noise exists on the loop from prior transmissions it is overwritten by the new transmission Nodes share the polling activity and any loss of polling is restarted automatically ACKs and NAKs are generated by end user nodes when they receive packets Each packet is tested against cyclic redundancy bytes in the packet A good check results in an ACK and a bad check results in a NAK Examples of use The ESM Multiloop Network is part of DCA s Hybrid Simulation Facility HSF The ESM provides DCA with a System Control Simulation Facility The uses of the system are to be outlined The various applications are in different stages of development some of the system uses are a direct result of the original implementation others require addi tional modeling application and demonstration software User ianguage The User Language provides the human interface to the system and demonstrates many of its modeling capabilities The User Language is an application program running on the various Host processors in the network All loop con nected terminals can commun
Download Pdf Manuals
Related Search