Performance Evaluation of Complex Systems: Techniques and Tools

Thefastevolutionandtheincreasedpervasivenessofcomputersandcom- nicationnetworkshaveledtothedevelopmentofalargevarietyofcomplex applicationsandserviceswhichhavebecomeanintegralpartofourdailylives. Modernsocietywidelyreliesoninformationtechnologies. Hence,theQuality ofService,thatis,thee?ciency,availability,reliability,andsecurityofthese technologies,isanessentialrequirementfortheproperfunctioningofmodern society. Inthisscenario,performanceevaluationplaysacentralrole. Performance evaluationhastoassessandpredicttheperformanceofhardwareandsoftware systems,andtoidentifyandpreventtheircurrentandfutureperformancebo- lenecks. Inthepastthirtyyears,manyperformanceevaluationtechniquesandtools havebeendevelopedandsuccessfullyappliedinstudiesdealingwiththec- ?gurationandcapacityplanningofexistingsystemsandwiththedesignand developmentofnewsystems. Recently,performanceevaluationtechniqueshave evolvedtocopewiththeincreasedcomplexityofthecurrentsystemsandtheir workloads. Manyoftheclassicaltechniqueshavebeenrevisitedinlightofthe recenttechnologicaladvances,andnoveltechniques,methods,andtoolshave beendeveloped.Thisbookisorganizedaroundasetofsurveypaperswhichprovideac- prehensiveoverviewofthetheories,techniques,andtoolsforperformanceand reliabilityevaluationofcurrentandnewemergingtechnologies. Thepapers,by leadinginternationalexpertsinthe?eldofperformanceevaluation,arebasedon thetutorialspresentedattheIFIPWG7. 3InternationalSymposiumonCom- terModeling,Measurement,andEvaluation(Performance2002)heldinRome onSeptember23-27,2002. Thepapersaddressthestateoftheartofthetheoreticalandmethodological advancesintheareaofperformanceandreliabilityevaluationaswellasnew perspectivesinthemajorapplicationdomains. Abroadspectrumoftopicsis coveredinthisbook. Modelingandveri?cationformalisms,solutionmethods, workloadcharacterization,andbenchmarkingareaddressedfromametho- logicalpointofview. Applicationsofperformanceandreliabilitytechniquesto variousdomains,suchas,hardwareandsoftwarearchitectures,wiredand- relessnetworks,Gridenvironments,Webservices,real-timevoiceandvideo applications,arealsoexamined. Thisbookisintendedtoserveasareferenceforstudents,scientists,and- gineersworkingintheareasofperformanceandreliabilityevaluation,hardware andsoftwaredesign,andcapacityplanning.VI Preface Finally,aseditorsofthebook,wewouldliketothankallauthorsfortheir valuablecontributionsandtheire?ortandcooperationinthepreparationof theirmanuscripts. July2002 MariaCarlaCalzarossa SalvatoreTucci TableofContents G-Networks:MultipleClassesofPositiveCustomers,Signals,and ProductFormResults ...1 ErolGelenbe SpectralExpansionSolutionsforMarkov-ModulatedQueues ...17 IsiMitrani M/G/1-TypeMarkovProcesses:ATutorial ...36 AlmaRiska,EvgeniaSmirni AnAlgorithmicApproachtoStochasticBounds ...64 J. M. Fourneau,N. Pekergin DynamicSchedulingviaPolymatroidOptimization...89 DavidD. Yao WorkloadModelingforPerformanceEvaluation...114 DrorG. Feitelson CapacityPlanningforWebServices(TechniquesandMethodology) ...142 VirgilioA. F. Almeida End-to-EndPerformanceofWebServices ...158 PaoloCremonesi,GiuseppeSerazzi Benchmarking...179 ReinholdWeicker BenchmarkingModelsandToolsforDistributedWeb-ServerSystems...208 MauroAndreolini,ValeriaCardellini,MicheleColajanni StochasticProcessAlgebra:FromanAlgebraicFormalismtoan ArchitecturalDescriptionLanguage ...236 MarcoBernardo,LorenzoDonatiello,PaoloCiancarini AutomatedPerformanceandDependabilityEvaluationUsingModel Checking ...261 ChristelBaier,BoudewijnHaverkort,HolgerHermanns, Joost-PieterKatoen Measurement-BasedAnalysisofSystemDependabilityUsingFault InjectionandFieldFailureData ...290 RavishankarK. Iyer,ZbigniewKalbarczyk VIII TableofContents SoftwareReliabilityandRejuvenation:ModelingandAnalysis...318 KishorS. Trivedi,KalyanaramanVaidyanathan PerformanceValidationofMobileSoftwareArchitectures ...346 VincenzoGrassi,VittorioCortellessa,Ra?aelaMirandola PerformanceIssuesofMultimediaApplications ...374 EdmundodeSouzaeSilva,RosaM. M. Le"ao,BerthierRibeiro-Neto, S'ergioCampos MarkovianModelingofRealDataTra?c:HeuristicPhaseTypeand MAPFittingofHeavyTailedandFractalLikeSamples...405 Andr'asHorv'ath,Mikl'osTelek OptimizationofBandwidthandEnergyConsumptioninWireless LocalAreaNetworks ...435 MarcoConti,EnricoGregori ServiceCentricComputing-27,2002. Thepapersaddressthestateoftheartofthetheoreticalandmethodological advancesintheareaofperformanceandreliabilityevaluationaswellasnew perspectivesinthemajorapplicationdomains. Abroadspectrumoftopicsis coveredinthisbook. Modelingandveri?cationformalisms,solutionmethods, workloadcharacterization,andbenchmarkingareaddressedfromametho- logicalpointofview.Applicationsofperformanceandreliabilitytechniquesto variousdomains,suchas,hardwareandsoftwarearchitectures,wiredand- relessnetworks,Gridenvironments,Webservices,real-timevoiceandvideo applications,arealsoexamined. Thisbookisintendedtoserveasareferenceforstudents,scientists,and- gineersworkingintheareasofperformanceandreliabilityevaluation,hardware andsoftwaredesign,andcapacityplanning. VI Preface Finally,aseditorsofthebook,wewouldliketothankallauthorsfortheir valuablecontributionsandtheire?ortandcooperationinthepreparationof theirmanuscripts. July2002 MariaCarlaCalzarossa SalvatoreTucci TableofContents G-Networks:MultipleClassesofPositiveCustomers,Signals,and ProductFormResults ...1 ErolGelenbe SpectralExpansionSolutionsforMarkov-ModulatedQueues ...17 IsiMitrani M/G/1-TypeMarkovProcesses:ATutorial ...36 AlmaRiska,EvgeniaSmirni AnAlgorithmicApproachtoStochasticBounds ...64 J. M. Fourneau,N. Pekergin DynamicSchedulingviaPolymatroidOptimization...89 DavidD. Yao WorkloadModelingforPerformanceEvaluation...114 DrorG. Feitelson CapacityPlanningforWebServices(TechniquesandMethodology) ...142 VirgilioA. F. Almeida End-to-EndPerformanceofWebServices ...158 PaoloCremonesi,GiuseppeSerazzi Benchmarking...179 ReinholdWeicker BenchmarkingModelsandToolsforDistributedWeb-ServerSystems...208 MauroAndreolini,ValeriaCardellini,MicheleColajanni StochasticProcessAlgebra:FromanAlgebraicFormalismtoan ArchitecturalDescriptionLanguage ...236 MarcoBernardo,LorenzoDonatiello,PaoloCiancarini AutomatedPerformanceandDependabilityEvaluationUsingModel Checking ...261 ChristelBaier,BoudewijnHaverkort,HolgerHermanns, Joost-PieterKatoen Measurement-BasedAnalysisofSystemDependabilityUsingFault InjectionandFieldFailureData ...290 RavishankarK. Iyer,ZbigniewKalbarczyk VIII TableofContents SoftwareReliabilityandRejuvenation:ModelingandAnalysis...318 KishorS. Trivedi,KalyanaramanVaidyanathan PerformanceValidationofMobileSoftwareArchitectures ...346 VincenzoGrassi,VittorioCortellessa,Ra?aelaMirandola PerformanceIssuesofMultimediaApplications ...374 EdmundodeSouzaeSilva,RosaM. M. Le"ao,BerthierRibeiro-Neto, S'ergioCampos MarkovianModelingofRealDataTra?c:HeuristicPhaseTypeand MAPFittingofHeavyTailedandFractalLikeSamples...405 Andr'asHorv'ath,Mikl'osTelek OptimizationofBandwidthandEnergyConsumptioninWireless LocalAreaNetworks ...435 MarcoConti,EnricoGregori ServiceCentricComputing-NextGenerationInternetComputing...463 JerryRolia,RichFriedrich,ChandrakantPatel EuropeanDataGridProject:ExperiencesofDeployingaLargeScale TestbedforE-scienceApplications ...480 FabrizioGagliardi,BobJones,MarioReale,StephenBurke AuthorIndex ...501 G-Networks: Multiple Classes of Positive Customers, Signals, and Product Form Results Erol Gelenbe SchoolofElectricalEngineeringandComputerScience UniversityofCentralFlorida Orlando,FL32816 erol@cs. ucf. edu Abstract. ThepurposeofthistutorialpresentationistointroduceG- Networks,orGelenbeNetworks,whichareproductformqueueingn- works which include normal or positive customers, as well as negative customers which destroy other customers, and triggers which displace othercustomersfromonequeuetoanother. Wederivethebalanceeq- tionsforthesemodelsinthecontextofmultiplecustomerclasses,show the product form results, and exhibit the tra?c equations which - in thiscase,contrarytoBCMPandJacksonnetworks-arenon-linear. This leadstointerestingissuesofexistenceanduniquenessofthesteady-state solution.GelenbeNetworkcanbeusedtomodellargescalecomputers- temsandnetworksinwhichsignalingfunctionsrepresentedbynegative customersandtriggersareusedtoachieve?owandcongestioncontrol. 1 Introduction In this survey and tutorial, we discuss a class of queueing networks, originally inspired by our work on neural networks, in which customers are either "signals" or positive customers.