Storage device performance prediction with CART models(17)

Storage device performance prediction is a key element of self-managed storage systems and application planning tasks, such as data assignment. This work explores the application of a machine learning tool, CART models, to storage device modeling. Our appr

[18]T.Joachims.Makinglarge-scaleSVMlearningpractical.InB.Sch¨olkopf,C.Burges,andSmola,

editors,AdvancesinKernelMethods–SupportVectorLearning.MIT-Press,1999.[19]K.Keeton,G.Alvarez,E.Riedel,andM.Uysal.CharacterizingI/O-intensiveworkloadsequentiality

onmoderndiskarrays.InTheFourthWorkshopOnComputerArchitectureEvaluationUsingCom-mercialWorkloads,2001.[20]ZacharyKurmas,KimberlyKeeton,andRalphBecker-Szendy.I/Oworkloadcharacterization.In4th

workshoponcomputerarchitectureevaluationusingcommercialworkloads,2001.[21]mb.Hardwarespendingsputters.RedHerring,pages32–33,2001.

[22]EdwardK.LeeandRandyH.Katz.Ananalyticperformancemodelofdiskarrays.InProceedingsof

the1993ACMSIGMETRICS,pages98–109,1993.[23]WillE.Leland,MuradS.Taqq,WalterWillinger,andDanielV.Wilson.Ontheself-similarnatureof

Ethernettraf c.InACMSIGCOMM,pages13–17,1993.[24]ArifMerchantandGuillermoA.Alvarez.DiskarraymodelsinMinerva.TechnicalReportHPL-2001-118,HPLaboratories,2001.[25]RuldolfH.Riedi,MatthewS.Crouse,VinayJ.Ribeiro,andRichardG.Baraniuk.Amultifrac-talwaveletmodelwithapplicationtonetworktraf c.IEEETransactionsonInformationTheory,45(3):992–1018,April1999.[26]B.D.Ripley.PaternrecognitionsandNeuralNetworks.CambridgeUniversityPress,1996.

[27]ChrisRuemmlerandJohnWilkes.Unixdiskaccesspatterns.InWinterUSENIXConference,pages

405–420,1993.[28]ChrisRuemmlerandJohnWilkes.Anintroductiontodiskdrivemodeling.IEEEComputer,27(3):17–

28,1994.[29]G.Seber.Multivariateobservations.Wiley,1984.

[30]ElizabethShriver,ArifMerchant,andJohnWilkes.Ananalyticalbehaviormodelfordiskdriveswith

readaheadcachesandrequestreordering.InProceedingsofInternationalConferenceonMeasurementandModelingofComputerSystems,pages182–191,1998.[31]MustafaUysal,GuillermoA.Alvarez,andArifMerchant.Amodular,analyticalthroughputmodel

formoderndiskarrays.InProceedingsof9thInternationalSymposiumonModeling,AnalysisandSimulationofComputerandTelecommunicationSystems,pages183–192,2001.[32]MengzhiWang,AnastassiaAilamaki,andChristosFaloutsos.Capturingthespatio-temporalbehavior

ofrealtraf cdata.PerformanceEvaluation,49(1/4):147–163,2002.[33]J.Wilkes.ThePantheonstorage-systemsimulator.TechnicalReportHPL–SSP–95–14,Hewlett-PackardLaboratories,1995.[34]JohnWilkes.TravelingtoRome,QoSspeci cationsforautomatedstoragesystemmanagement.In

Proc.Intl.WorkshoponQualityofService(IWQoS’2001),pages75–91.Springer-VerlagLectureNotesinComputerScience2092,2001.[35]JohnWilkes.Dataservices–fromdatatocontainers.KeynoteaddressatFileandStorageTechnologies

Conference(FAST’03),March–April2003.