ALEVIN Software User Guide
Contents
1. Add Demand to VirtualNode 1 1 Select the Demand to add CpuDemand demanded cycles Add Demand Cancel and set the parameter s for the selected constraint A constraint is required for the node to be created After following this steps the node will be created at the desired position and Shown in the graph panel Creating Links To create a link select two nodes and right click on the layer they belong to You then have to choose source and destination of the link to be created To make things simple the two alternatives are displayed in the pop up menu Add link SubstrateNWode 16 SubstrateNode 1 e Add link SubstrateNode 1 gt SubstrateNode 16 After the desired link is selected from the pop up menu you need to add a constraint to the link for it to be created For this purpose a dialog as described in the Creating Nodes section will be displayed As in the case of nodes a constraint is required for a link to be created After adding a constraint the link will be created and shown in the graph panel 3 1 3 Editing Nodes and Links To edit a node or a link right click on it You have the following options Add Constraint lf you choose to add a constraint you will need to select the constraint type to add and set it s parameter s as described in the Creating Nodes section Note that each node link can only have a single resource of a given type Edit Constraint If you select the edit constraint2. Submit Cancel k Is the number of shortest paths in terms of hops used in the algorithms for the virtual link mapping stage in case that rounded paths do not accomplish the virtual link demands Cpu and Bandwidth Weigths This values allow to give more revenue value to Cpu or Bw when the virtual network embedding is being performed i e if Cpu weight is higher than bandwidth weight the algorithm will perform a mapping trying to minimize the CPU consume in the substrate network Distance Considered It refers to the maximum distance that a substrate node can be from a virtual node to be considered a candidate node to be mapped This value must be between O and 100 the area of the space where each network is plotted is 100x100 Node Overload The node overload is not a value presented in the original algorithm Current algorithms do not allow that more than one virtual node belonging to a virtual network are mapped in one substrate node This value modifies the algorithm to avoid this constraint Again If the No option is chosen the algorithm will have its normal behavior Algorithm Type This algorithm work in the node mapping phase by solving a relaxed version of the NP Complete Unsplittable Flow Problem UFP After that the relaxed solution is rounded by two method deterministic or randomized This parameter contains the rounding type Hidden Hops Selection In the next step you can select the hidden hop dema
3. the relaxed solution is rounded by two method deterministic or randomized This parameter contains the rounding type To see more details of previous two algorithms please refer to INPROCEEDINGS CHO9 author Chowdhury N M M K and Rahman M R and Boutaba R booktitle INFOCOM 2009 IEEE title Virtual Network Embedding with Coordinated Node and Link Mapping year 2009 month april volume number pages 783 791 keywords Internet coordinated node heuristic based algorithms link Mapping mixed integer program multiple heterogeneous virtual networks network virtualization substrate network augmentation virtual network embedding Internet embedded systems virtual machines doi 10 1109 INFCOM 2009 5061987 ISSN 0743 166X e Coordinated Node Node Mapping and Rounding Multipath Link Mapping This solution uses the heuristic solution of the Mixed Integer Programming approach in the node mapping phase and a new approach in link mapping to realize the virtual link mapping using single path Single path is reached by rounding the multi path solution in the same way than in the previous proposal Deterministically or randomly Setting Parameters Wizard Parameters Maximum distance gt E gt RIJ kp CPU weight Bandwith weight Number of k Type of algorithm Deterministic Ramdomized ls node overload considered very O No
4. Scenario Export To export the current scenario select Export from the File Menu or use the Ctrl S hotkey You will then need to choose the file to export to and select Save The Scenario including constraints and mappings will be saved as XML data to the file specified 3 1 5 Scenario Generation Network Topology Generation ALEVIN provides a random scenario generator the scenario wizard For the scenario generation the number of nodes in each network is set and the links are added randomly using a Waxman generator It creates edges within a 1x1 Square with probability where is the Euclidian distance between vertex and vertex and is the maximum distance between any two nodes An increase in the parameter alpha increases the probability of edges between any nodes in the graph while an increase in beta yields a larger ratio of long edges to short edges For more information on the waxman generator please refer to article Waxman88 author Waxman Bernard M title Routing of Multipoint Connections journal IEEE Journal of Selected Areas in Communication pages 1617 1622 number 9 volume 6 month dec year 1988 Prior to setting any parameters the scenario wizard looks as follows Scenano Wizard Substrate Network Substrate nodes Alpha Beta Virtual Networks number 15 default number of virtual nodes i nodes alpha 11 1 0 To generate a new scenari
5. is shown Progress Processed VN links 100 Mapped VN links 100 Algorithms progress dialog 4 2 2 Implemented Algorithms Until now there are five algorithms available Simple Dijkstra Test virtual link mapping algorithm with node mapping assumed as performed mapping each virtual link to the shortest path Greedy Available with K Shortest Path Virtual node and link mappings for each VNR are performed separately Node mapping is performed using a greedy algorithm that maps virtual nodes with higher revenues to SN nodes with higher available resources GAS Link mapping is accomplished by mapping each virtual link to the shortest path accomplishing capacity constraints between the corresponding mapped nodes in the SN Greedy Available with Path Splitting Virtual node and link mappings for each VNR are performed separately Node mapping is also performed witha greedy available resources GAS Link mapping is accomplished with a multi path MP optimal solution that avoids the NP completeness of the problem Coordinated Node and link mapping with Path Splitting The node mapping Stage is performed by defining an augmented graph over the substrate network introducing a set of meta nodes one per virtual node each connected to a cluster of candidate SN nodes obeying location and capacity constraints The algorithm solves the VNE problem by using a Mixed Integer Programming MIP Its objective is to minimize the
6. layers from the File Menu or use the Ctrl N Hotkey In the next step you will be promted for the number of virtual networks of the scenario to be created 2 Number of virtual networks Abbrechen A scenario consisting of a substrate network and the desired number of virtual networks is created and shown in the graph panel All created layers substrate or virtual will be empty To add nodes and links follow the steps described in the Scenario Editing section below 3 1 2 Scenario Editing Adding Networks To add virtual networks select New empty scenario networks from the File Menu Next you will be promted for the number of virtual networks to add Q m Number of virtual networks Abbrechen Substrate Network Virtual Network 1 Virtual Network 2 4 j 4 j k Virtual Network 3 Virtual Network 4 ah ae 4 t 4 t The desired number of virtual networks is added to the graph If we add two more networks to our exemplary scenario the graph will be as follows Note that the new networks are empty Creating Nodes A node can be created in an existing network Substrate or virtual network at a desired position To create a node right click at the desired position and select Create Node from the pop up menu Create node Maximize Hide Next you need to add a constraint to the node to be created Select the Demand to add select Demand select Demand CpuDemand
7. the following k Is the number of shortest paths in terms of hops used in the algorithms for the virtual link mapping stage Cpu and Bandwidth Weigths This values allow to give more revenue value to Cpu or Bw when the virtual network embedding is being performed i e if Cpu weight is higher than bandwidth weight the algorithm will perform a mapping trying to minimize the CPU consume in the substrate network Distance Considered It refers to the maximum distance that a substrate node can be from a virtual node to be considered a candidate node to be mapped This value must be between O and 100 the area of the space where each network is plotted is 100x100 Node Overload The node overload is not a value presented in the original algorithm Current algorithms do not allow that more than one virtual node belonging to a virtual network are mapped in one substrate node This value modifies the algorithm to avoid this constraint Again If the No option is chosen the algorithm will have its normal behavior Algorithm Type This algorithm work in the node mapping phase by solving a relaxed version of the NP Complete Unsplittable Flow Problem UFP After that the relaxed solution is rounded by two method deterministic or randomized This parameter contains the rounding type e Coordinated Node Node Mapping and Link Mapping with Path Splitting Link Mapping This solution uses the heuristic solution of the Mixed Integer P
8. the substrate and arbitrary virtual networks as directed graphs The next figure depicts the architecture of ALEVIN and highlights the modular interaction of parameters for substrate as well as virtual networks algorithms and evaluation SubstrateNetwork l NetworkStack L VirtualNetwork shows modifies based on consists of MuLaVilo JO S SISTIOD shows Mapping calculates evaluates ALEVIN provides the ability to illustrate the deployment of resources in the Substrate network and demands in an arbitrary number of virtual networks as well as the mapping of demands on resources calculated by a VNE algorithm Moreover ALEVIN can be used to create VNE scenarios as well as import and export them using an XML based exchange format ALEVIN is completely modular regarding the addition of new parameters to the VNE model Using the java visitor design pattern in a sophisticated way we are able to avoid any casts to concrete demand resource classes Thus the number of parameters is not performance relevant and a convenient implementation of arbitrary parameters is possible To increase ALEVIN s modularity and to make it a flexible and extensible platform to compare existing and upcoming algorithms the implementation of algorithms is kept independent of the resource demand implementation To that end a simple interface is provided defining the rough structure of an algorithm and connecting its output to th
9. to minimize the CPU consume in the substrate network Distance Considered It has the same meaning than in the previous algorithm Node Overload It has the same meaning than in the previous algorithm Previous two algorithms are proposed in article Yu08 o E _ author Yu Minlan and Yi Yung and Rexford Jennifer and Chiang Mung title Rethinking virtual network embedding substrate support for path Splitting and migration journal SIGCOMM Comput Commun Rev volume 38 issue 2 month March year 2008 issn 0146 4833 pages 1 29 numpages 13 url http do1l acm org 10 1145 1355734 1355737 doi http doi acm org 10 1145 1355734 1355737 acmid 1355737 publisher ACM address New York NY USA keywords network virtualization optimization path migration path Splitting virtual network embedding e Coordinated Node Node Mapping and Link Mapping with k Shortest Paths Link Mapping This solution uses the heuristic solution of the Mixed Integer Programming approach in the node mapping phase and the k Shortest path approach in the link mapping phase Setting Parameters Wizard x Parameters Maximum distance CPU weight Bandwith weight Number of k Type of algorithm Pipers O Ramdomized ls node overload considered O Yes a No Submit Cancel The parameters presented in the algorithm s wizard are
10. ALEVIN Software User Guide Authors Michael Duelli and Daniel Schlosser University of Wuerzburg Juan Felipe Botero and Xavier Hesselbach Universitat de Catalunya Andreas Fischer University of Passau Table of Contents TI CFOGUGCEION iaxsaadescieanaaaisaevsonne tysnsascohmcanmndiinesmciabi wana AE ET 2 ZINSEAlLatiOn REGUIFGMCNUS waiisasiascivecetedbbimiraaninae nadine biniieeinaeoninninens 3 2 1GLPK LP solver framework INStallation cccccececeseeeeeeeeeeeeeeeeeeneaneaneas 3 2 1 1Unpacking the IStriDUTION TFIl ec ceeee eee ee eee eeeeeeeeeesneateateaneass 4 2 1 2Configuring TNE PACKAGE cccceececeeeeeeeeeeeeeseeeeaeeseeaeeaeeaneateaneaneass 4 7 LSCOMPINNG THE PACKI E tiiaiccascianstiaiccennnsaravetnonannsamienieennias Aiaaninearbebinins 4 2 1 4Check the packager nisoria a a aa aE NA TA 4 2 1 5lnstall the package sssssessrrssrrrsrrrrsrrrrrrrrsrrrrsrrrerrrrsrrrrrrresrerrrrnerrre 4 SEEN e I E 6 3 1Create Edit Import Export and Manage ScenarioS sssssssrsssesrererreres 6 3 lL TS enanie Creation From Scrat Noir a a 6 3 1 29ce an o EQN rnas a a E a A 6 3 LASEGITING NODES ANA LINKS aeania n AE A EEA 8 3 4 SCenanlo Import and EX PO Gacr E 9 LISEN Genera Oena ENAA 10 ACUrFENLIY S PDOtEdAIJOritNM S eiaa T Aan 15 4 lAlgorithms brief description esssssessesseresneesennesrensennesennnnnnnssennnenesseernne 15 4 1 1Virtual Node Mapping AlgorithMS ssssessssssrressrnssrrrsrrreser
11. cost i e the weighted sum of the bandwidth and CPU allocated in the SN links to fulfill VNR demands of embedding a VNR To avoid the NP completeness of the MIP its linear programming relaxation is solved and the obtained solution is rounded in two ways deterministically or randomly The link mapping stage is performed using a Multi Path approach MP Coordinated Node and link mapping with k Shortest Path The same virtual node mapping stage is performed in the same way MIP as the previously described algorithms The link mapping stage is performed using a K Shortest paths approach Input Parameters Each algorithm has different input parameters For the implemented algorithms the parameters are as follows Greedy Available Resources Node Mapping k Shortest paths Link Mapping The only parameter needed in this algorithm is the number of shortest paths to be calculated the parameter k An input wizard is shown Setting Parameters Wizard Parameters Number of k Is distance considered Maximum distance ls node overload considered Submit Cancel The parameters presented in the algorithm s wizard are the following k Is the number of shortest paths in terms of hops used in the algorithms for the virtual link mapping stage Distance Considered The default value of this boolean value is No This is not a value presented in the original algorithm This value is included to provide the po
12. e GUI as illustrated in the previous picture 2 Installation Requirements Being implemented in Java Alevin is platform independent and should work on MS Windows Mac OS X and Linux For proper operation of Alevin however the following software packages have to be installed e SUN Java JDK version 1 5 or later recommended version 6 update 21 or later Do not use other Runtime Environments as they are known to create problems e The GNU Linear Programming Kit GLPK http www gnu org software glpk this is used by some algorithms 2 1 GLPK LP solver framework installation The GLPK GNU Linear Programming Kit package is intended for solving large scale linear programming LP mixed integer programming MIP and other related problems It is a set of routines written in ANSI C and organized in the form of a callable library http www gnu org software glpk Project GLPK for Java delivers a Java language binding http sourceforge net projects glok java Some algorithms of ALEVIN use linear programming to realize the VNE ALEVIN implementation relies on GLPK to solve the VNE node and link mapping stages when they require Linear Programming As ALEVIN is developed in java we have use the java biding interface provided in http sourceforge net projects glpk java The installation instructions are placed in the doc folder of the glpk java software The file glok java pdf explains how the installation mus
13. ed Integer Programming Solution Substrate Network Graph is augmented creating Meta nodes representing the virtual nodes and meta edges joining meta nodes with the candidate nodes to be mapped in the SN they are added to the SN graph as indicated in Figure 4 Meta edge gt Meta node CC Cluster ws 5 Virtual Network Augmented Graph Figure 4 In figure 4 a will be mapped to either A or C A and C are the candidates nodes to map a After the following steps the map is decided 1 The problem is solved using linear programming LP gt The relaxed BIP formulation is solved 2 The LP solution contains rational value for each of the meta edges joining the meta nodes with nodes in the substrate network A randomized or deterministic rounding is performed among the meta edges of each meta node to choose one of them the SN connected to the chosen meta edge is then chosen as the mapped node in the SN 4 1 2 Virtual Link Mapping Algorithms It is very important to state that the virtual link mapping stage starts after virtual node mapping K Shortest Paths Algorithm KSP This algorithm calculates for each virtual link the k shortest paths from one mapped virtual node to the other end in the substrate network The mapping is performed by assigning to each virtual link the first shortest path that accomplishes the bandwidth and CPU demands Multi Path Algorithm MP This algorithm solve
14. f Request Node Stress In a node i V the node stress Sn i is the number of virtual nodes mapped on top of 1 Link Stress In a link i j E A the link stress S 1 is the number of virtual links using i j as part of its mapped A SN path B Substrate Network Figure 2 The numbers above the links an nodes are the respective stresses The objective of this approach is to minimize the balanced stress weigthed sum of node and link stresses in the network To reach this objective the virtual node mapping stage is performed by assigning the nodes with greater degree in virtual network to the nodes with greater potential for each node in the SN the potential is the multiplication of the node stress by the sum of adjacent link stresses in the substrate network Greedy Available Resources GAS This virtual node mapping approach is very similar to the GS The available resources concept is defined in Figure 3 CONCEPTS DEFINED IN EXISTING APPROACHES Available Resources H Is a substrate node parameter It measures node capacity taking into account the capacity of its incident links H l NC X LCi 5 LEV i j EA i j EL L Available resources Substrate Link Capacity 5 5 9 70 i Substrate Node A node A H A Figure 3 The virtual node mapping GAS is a greedy algorithm Nodes with greater demand are assigned with greater available resources Mix
15. he selection panel shows the network entities currently selected as well as their constraints in a tree structure Here is an example using our exemplary scenario Note that some elements of the selection tree are not shown because their parent element are not popped up Substrate Network Virtual Network 1 Virtual Network 2 oo 4 j 4 F 4 j _ Selection panel example 5 3 The Mapping Panel Selection Mapping JE Nodes 5 hd SubstrateNode 3 pe IdResource id 3 i i CpuResource cydes 100 SubstrateNode 2 ER ea VirtualNode 13 1 pe VirtualNode 17 2 ee IdDemand demanded id 2 i CpuDemand demanded cydes 10 ce J VirtualNode 16 2 i Links 2 SubstrateLink 9 i dj BandwidthResource bandwidth 10Mbit s Gi VirtualLink 18 2 i BandwidthDemand demanded bandwidth 5 Mbit s The mapping panel shows the mappings of the current scenrio in a tree Sstrucure If no mappings exist the virtual links will be the leaf elements of the tree Selection Mapping W Virtual Networks Fh di Virtual Network 1 E d Virtual Network 2 4 Substrate Network Mapping panel showing only the virtual links as no Mappings exist lf mappings do exist they are shown as child elements of the virtual links When selecting a virtual link from the mappings tree it is highlighted along with the substrate links it is mapped to Hidden hops are displayed using a diffe
16. ion on the current scenario select it from the menu The result will be displayed in a dialog that also enables copying it An exaple for the average node stress is shown below al AverageNodeStress Evaluation result 0 8 Cer Co Metrics Dialog
17. jects glpk java 2 1 1 Unpacking the distribution file Copy the distribution file to a working directory Check the MD5 checksum with the following command md5sum glpk java X Y tar gz Unpack the archive with the following command tar xzf glpk java X Y tar gz Now change to the new direcotry glpk java X Y 2 1 2 Configuring the package Open swig Makefile in a text editor e g with the following command vi swig Makefile Adjust the installation path prefix Adjust the include path for glpk h Adjust the version information concerning GLPK Save the text file On Mac OS X jni h is in the following path System Library Frameworks JavaVM f ramework Headers 2 1 3 Compiling the package To remove all files from prior compiling use the command make clean The package is compiled with the command make 2 1 4 Check the package To check if everything is built correctly use the command make check 2 1 5 Install the package To install the package you must be root or a suodoer As sudoer use the command sudo make install 1 Mac Os The process to install GLPK java in MacOS is the same as it is in linux Remember that on Mac OS X jni h is in the following path System Library Frameworks JavaVM f ramework Headers 3 Editing 3 1 Create Edit Import Export and Manage Scenarios 3 1 1 Scenario Creation From Scratch To create a new empty scenario select New empty scenario
18. leared or redirected to the normal terminal console It can also be closed and then restored by using the Hidden Panels option in the Views Menu 5 5 Menu Options 5 5 1 File Menu aa New empty scenario layers Ctrl N E Import Ctrl 0 ka Export Ctrl 5 Close Ctrl W a Quit Ctrl Q File Menu From the File Menu you can select one of the following options e Creating a new empty scenario or adding new layers to an existing scenario e Importing a scenario e Exporting a scenario e Closing the Scenario e Exiting the Application 5 5 2 View Menu Show node labels Show link labels Hidden Panels view menu without hidden panels You can check the Show node labels or Show link labels options This will display the labels of nodes links in the graph panel The Hidden Panels menu entry provides a means to restore closed or hidden GUI components Currently only the console can be closed and restored This functionality is provided by MuLaViTo and is only available if there are closed GUI components available for restoration Show node labels Show link labels Hidden Panels Console view menu with hidden console 5 5 3 Generators Menu Scenario Wizard Generate Constraints Clear all constraints Generators Menu The Generators Menu provides the following three options e Generate a scenario using the Scenario Wizard e Generate constraints for an existing scenario e Remove all constraints
19. nds that are to be considered when computing the mapping This is done using the Hidden Hops Dialog Set HiddenHops oo m Hidden Hops Hidden Hop Demand Factor BandwidthCpuHiddenHop Set Cancel Hidden Hops Dialog It consists of a table that enables you to select the HH Demands to use and set the factor used for computing them Output Parameters After termination some information messages are displayed in the Console If the algorithm finishes normally the computed mappings are added to the scenario and displayed in the Mapping Panel Remove All Mappings To remove all mappings of the current scenario select the Remove all mappings option from the Algorithms Menu This action will also free up all resources in the substrate network 5 GUI Features 5 1 The Graph Panel Substrate Network Virtual Network 1 Virtual Network 2 i i i F J F a Ir af F The graph panel showing our exemplary scenario The graph panel is taken from MuLaViTo and is the central part of the GUI where the scenarios are visualized For each layer of the scenario substrate or virtual the graph of the network is displayed The graph panel enables you to 5 1 1 Move Nodes To move nodes simply drag them with the mouse Note that this action will have an effect on the node s coordinates 5 1 2 Zoom in and out To zoom in or out hold the Ctrl key and use the mouse wheel 5 2 The Selection Panel T
20. o the following parameters must be set e For the substrate network e The number of substrate nodes e The value of the alpha and beta parameters e The number of virtual networks e The default number of virtual nodes per network e For each virtual network e The number of nodes if different from the default e The value of the alpha and beta parameters To set parameters for a specific virtual network just click on the respective table row If the mouse stands still over one of the table columns a short info about the parameter is displayed as shown in the graphic below After setting the parameters as needed the aspect of the scenario wizard changes slightly Scenano Wizard j ES Substrate Network Alpha Beta Virtual Networks number 94 2 default number of virtual nodes alpha gt 0 An increase in alpha will increase the number of edge Constraints Generation Generating constraints is possible using the constraints generator Generate Constraints timm Resource Maximum parameter values BandwidthResource max bandwidth Maximum parameter values Constraints Generator It consists of two sections one for the resources and one for the demands The most important component of the constraints generator is the table used for selecting the constraints to add and setting the required parameters The resources panel contains one such table while the demands panel cosists
21. of one table for each virtual network the tables being packed in different tabs For constraints generation a constraint needs to be selected for the desired network substrate or virutal and the corresponding parameters set if needed This is done using the tables described above Currently only CPU and bandwidth constraints have parameters to set All other constraints need no further user action to be generated correctly Demand genration can be done individually on a per virtual network basis After selecting some constraints and setting the needed parameters the constraints generator will look as seen below Notice that as described above setting additional parameters is not required for all resource demand types Genere conssim E Maximum parameter values max bandwidth a me OO Ta beo lt Constraints Generator with selected constraints and parameter set Removing All Constraints It is also possible to remove all constraints of the current scenario for example in order to regenerate them 4 Currently Supported Algorithms 4 1 Algorithms brief description In this section a brief description of the virtual node and link mapping existing algorithms is given The explanation will separate the virtual node mapping and link mapping 4 1 1 Virtual Node Mapping Algorithms Greedy Stress Approach GS To understand the objective of this approach Figure 2 defines the term stress oF ual Network T tual twork H o
22. of the current scenario 5 5 4 Algorithms Menu Simple Dijkstra Greedy Available and K Shortest Paths Path Splitting Coordinated Node and Link Mapping with Path Sppliting Coordinated Node and Link Mapping with k Shortest Paths Coordinated Node Mapping with Rounding PathStripping Link Mapping Basic VN Basic VN Link Stress Subgraph lemerphism Advanced Subgraph lomerphism Optimal Mappings Algorithm Remove all mappings Algorithms Menu From the algorithm menu you can either select an algorithm to be rum or remove all mappings of the current scenario Run an Algorithm You can run an algorithm in order to create mappings between demands and resources See the graphics above for a list of the currently available algorithms An algorithm can only be run if the current scenario has no mappings Remove all Mappings This option removes all existing mappings of the current scenario and frees up all resources in the substrate network 5 5 5 Metrics Menu AcceptedVnrRatio AvActiveNodeStress AvNodeStress Cost CostRev TimesMappedRev CostRevenue LinkCostPerVnr MappedRevenue RatioMappedRevenue ReyectedNetworksNumber RemainingLinkResource RevenueCost Running Time TotalRevenue Metrics Menu The Metrics Menu enables the evaluation of the current scenario using one of the available metrics It is generated automaticly using reflection to ensure that all available eviluations are displayed To perform an evaluat
23. option you can edit the parameter s of the node s link s constraints First you need to select the constraint to be edited Edit Resource of SubstrateNode 0 5 then you can set the new values of the parameter s Edit Resource of SubstrateNodel0 25 select the Resource to edit CpuResource cydes 100 Edit Resource Cancel Remove Constraint To remove a constraint select it as described in the Edit Constraint section above and click on the Remove button Delete Node Link If you select this option the network entity along with its constraints and mappings will be deleted If you delete a node all its incident links will be deleted as well 3 1 4 Scenario Import and Export Scenario Import TO import a scenario select Import from the File Menu or use the Ctrl O Hotkey Next select the file to import from Scenario Import Suchen in di XML L 1 svn tlhe Do Examplel Zuletzt Example verwendet Desktop exemplary scenario mappings exemplary scenario F Eigene Dokumente A Computer Ta Dateiname test xml Netzwerk rune Dateityp All supported formats Import file chooser Substrate Network Virtual Network 1 Virtual Network 2 pe k ai k k F b Fi H Pi I Exemplary scenario import The imported scenario will then be displayed in the graph panel of the GUI For our exemplary scenario this looks as follows
24. rent color magenta Substrate Network Virtual Network 1 Virtual Network 2 Selection Mapping 2 A E Virtual Networks E i Virtual Network 1 E a TTT TTN ee VN 13 gt IdDemand demanded z VN 13 gt CpuDemand demand ve VL 15 gt BandwidthDemand de A e VL 15 gt BandwidthDemand de re we VL 15 gt BandwidthDemand da a T E VL 15 gt CpuDemand demand T a 1 0 VL 15 gt CpuDemand demand a E E VN 14 gt IdDemand demanded Set VN 14 gt CpuDemand demand Virtual Network 2 J Substrate Network 4 4 t t Mapping panel with mappings and highlighting Note that the Virtual Network 2 node is not popped up in this example The Mapping panel also enables reverse highlighting If you select a substrate link the virtual network entities mapped on the link or its incident nodes are highlighted Substrate Network Virtual Network 1 Virtual Network 2 Mapping Ee Virtual Networks Virtual Network 1 4 VN 13 gt VL 15 gt VN 14 Virtual Network 2 Substrate Network D SN 1 gt SL 7 gt 5N 2 SN 5 gt SL 8 gt SN 1 i SN 2 gt SL 10 gt 5N 4 SN 4 gt SL 11 gt 5N 3 S5N 4 gt S51 12 gt SN 5 I HAH ARE ees h i 4 j 4 4 p Mapping panel with mappings and highlighting 5 4 The Console The Console is taken from MuLaViTo and displays information debug and error messages The Console can be c
25. rogramming approach in the node mapping phase and a path splitting approach in the link mapping phase Setting Parameters Wizard Parameters Maximum ae CPU weight Bandwith weight Type of algorithm Deterministic O Ramdomized ls node overload considered Yes O No Submit Cancel The parameters presented in the algorithm s wizard are the following Cpu and Bandwidth Weigths This values allow to give more revenue value to Cpu or Bw when the virtual network embedding is being performed i e if Cpu weight is higher than bandwidth weight the algorithm will perform a mapping trying to minimize the CPU consume in the substrate network Distance Considered It refers to the maximum distance that a substrate node can be from a virtual node to be considered a candidate node to be mapped This value must be between O and 100 the area of the space where each network is plotted is 100x100 Node Overload The node overload is not a value presented in the original algorithm Current algorithms do not allow that more than one virtual node belonging to a virtual network are mapped in one substrate node This value modifies the algorithm to avoid this constraint Again If the No option is chosen the algorithm will have its normal behavior Algorithm Type This algorithm work in the node mapping phase by solving a relaxed version of the NP Complete Unsplittable Flow Problem UFP After that
26. rrerssrrere 15 4 1 2Virtual Link Mapping AIQOrithims ccccccceceseeeeeeeeeeeeeeeeaeeesneeneneas 16 4 2Algorithms and MappingS sssssssssssssrrssrrrsrrrserrrsrrrrrrrrserrrsrrrerrrrerrrrerrn gt 17 Ae aD DINO ara A AA 17 4 2 2Implemented AIQOriths ccccceceeeeceeeeeeeeeeeeeseeeesesneseeaneaeeaneaneas 18 SGU FeatUr C eaea a a era unten buniakeedebageaatads 25 S LTE GALON PING heraa a E AA 25 S Le MOVE NOAGS ascites ints acne inden a A T N AE 25 Dla ZZOOM WIM AMO OUE a a a a acnnnns bane 25 S2 LINE SCICECTION PAN Clisiciitatisahiitrii titre niente TA 25 Do TNE MapN Pa Me leeraar 26 Di AINE CONS OG pan nss sna vivennncnasscinsenmweteteatntaniberndane r EAEAN 2 SOMEN ODUONS ariran A AE EA E exssauidanatagiian 28 S LPIS MeN U ap N T Sita 28 De DZ VIO MON Ue aiicwecaasinoauwiiene euk NA AAU eE EE aAA A 28 JAGENE MONU a teeenenormeeniwoeaaienicesaeieseiae 29 SO AAO AMS MeN Usein aaa 29 S39 5Metric Ss MeN enea a a A ENE 30 1 Introduction The focus in the development of ALEVIN is on modularity and efficient handling of arbitrary parameters for resources and demands and the support of integration of new and existing virtual network embedding algorithms and evaluation metrics For platform independency ALEVIN is written in Java ALEVIN s graphical user interface GUI and multi layer visualization component is based on the MuLaViTo project MuLaViTo http mulavito sf net which enables us to visualize and handle
27. s a LP formulation of the virtual link mapping stage the Same formulation as the relaxed BIP This formulation is equivalent to the multi commodity flow problem and is solved using the traditional LP methods SIMPLEX IPM etc A example of this approach is shown in Figure 6 O Virtual Network C E Substrate ae Network a Path 1 Path 2 Figure 6 Rounding Multi Path RMP The problem is formulated as the multi commodity flow problem and is solved using optimal linear programming algorithms SIMPLEX IPM etc Then the multi path solution each path will have a rational percentage of the demand is rounded in a deterministic or randomized way and just one directed path is used to map each virtual link Subgraph isomorphism detection heuristic SID The mapping in nodes and links is done simultaneously by trying to find a Subgraph inside the substrate network being isomorphic to the virtual network request Graph isomorphism is explained in figure 5 4 2 Algorithms and Mappings 4 2 1 Mappings Compute Mappings To compute mappings for the current scenario you need to run an algorithm from the Algorithms Menu To do so just select it from the menu Algorithms can not be run on scenarios that already have mappings To better understand the input parameters requested by each algorithm see a brief and detailed explanation of the algorithms While an algorithm is running a progress dialog
28. ssibility of comparing this algorithm with subsequent proposals lf the value is chosen to No the normal behavior of the algorithm is presented The value refers to the maximum distance that a substrate node can be from a virtual node to be considered a candidate node to be mapped This value must be between O and 100 the area of the space where each network is plotted is 100x100 Node Overload As the previous parameter the node overload is not a value presented in the original algorithm Current algorithms do not allow that more than one virtual node belonging to a virtual network are mapped in one Substrate node This value modifies the algorithm to avoid this constraint Again If the No option is chosen the algorithm will have its normal behavior e Greedy Available Resources with Path Splitting Link Mapping To realize the link mapping algorithm it is needed to provide the weight that user gives to the CPU and BW Setting Parameters Wizard Parameters CPU weight Bandwith weight Is distance considered Maximum distance l Is node overload considered Yes O No Submit Cancel The parameters presented in the algorithm s wizard are the following Cpu and Bandwidth Weigths This values allow to give more revenue value to Cpu or Bw when the virtual network embedding is being performed i e if Cpu weight is higher than bandwidth weight the algorithm will perform a mapping trying
29. t be done However here a step by step explanation will be provided for each platform 1 Windows The GLPK for Java JNI library can be compiled from source code The build and make files are in directory w32 for 32 bit Windows and in w64 for 64 bit Windows The name of the created library is glok 4 45 java dll for revision 4 45 Recommended A precompiled version of GLPK for Java is provided at http winglpk sourceforge net The library the name of the library is glpk 4 45 java dll has to be in the search path for binaries it can be found in the w64 or w32 folder depending of your computer architecture of the percompiled version http winglpk sourceforge net Either copy the library to a directory that is already in the path e g C windows system32 or update the path in the system settings of Windows The jar file needed also for the execution of ALEVIN is stored also in the same w64 or w32 folder The library has to be in the CLASSPATH Update the classpath in the system settings of Windows or specify the classpath upon invocation of the application e g java classpath glpk java jar MyApplication l1 Linux To install the LINUX version the original GLPK library http www gnu org software glok must be compiled in first place The GLPK for Java JNI library can be compiled from source code The following instructions are also contained in the file INSTALL provided in the source distribution http sourceforge net pro
Download Pdf Manuals
Related Search