国立台北大学第十二次校务会议

中国科学技术大学硕士学位论文云环境下多维属性驱动的资源可用性研究姓名：是丽俐申请学位级别：硕士专业：计算机系统结构指导教师：杨寿保2011-04I()SOAQoS1.2.3.CPU4.IIAbstractIIIAbstractWiththedevelopmentofnetworkbandwidth,theconditionofaccessingremoteservice(forexample,dataprocessing,storageandinformationservice)throughnetworkhasbecomemoreandmoremature,andthencloudcomputingappears.Itintegratesgridcomputing,virtualizationandSOAetc.togetherwhilegatheringstorage,platform,applicationandcomputingcapabilitywhichisabstract,virtualandscalableintoresourcepool,toprovideservicetousersaccordingtotheirneeds.Usershavenoknowledgeoftechnologiesdetailsatall,buttoaccessthecloudthroughinterfaces.Ascloudenvironmentishighlycommercializedanduser-central,itisimportantforresearcherstosolvetheproblemofprovidingserviceofhighscalability,highavailability,highsecurityandhighefficiency.However,atthesametime,thecomplexityofcloudisblockingintheway.ResourceManagementisthecorecomponentincloudenvironment.Thegoalofresourcemanagementincloudistomakeuseofresourceseffectivelyandreasonably,whichcanacquirehighquality,efficiencyandprofitatlesscost.Therearetwomajormanagementdifficultieshere:oneishowtomakefulluseofhistoricandcurrentinformation,toachieveeffectiveresourceschedulingtomeettheperformanceandQoSneedsofusers;theotherisdatamanagementensuringhighavailabilityofdataaccessserviceatthepremiseofgainingefficiencyandconsistency.Focusingonthesetwoissuesofresourceschedulinganddatamanagement,thisthesiscarriesoutresearchesasfollows:1.Investigatetheresearchbackground,includingtheconcept,featuresandapplications.Statetheprocessandbackgroundofcloudcomputing.Analyzethedifferencesandconnectionsbetweengridcomputingandcloudcomputing.Thensimplyintroducecurrentrelatedresearchesandproposetheresearchtopics.2.Introducethecontentofresourcemanagementincloudenvironment.Summarizetheresourcetypesandanalyzethekeytechniquesofresourcemanagementwhichinvolveresourcescheduling,datamanagementandvirtualmachine,energyconsumptionmanagementetc.Thendiscusstheproblemsfacinginfrontofthetechniques.3.ComprehensivelyconsidertherateofCPUusage,levelofnetworkload,andAbstractIVresourcefailuretoforecastresourcefuturestateforgettingbetterjobschedulingresults.Anevaluationmeasurementisdesignedtoquantifythepredictionresultforscheduling.Experimentsshowthatthismethodhasabetterperformanceonresourceidlerateandpredictionaccuracy.4.Availabilityisthekeyelementincloudstorage.Thecomplexityofcloudstoragebringsavailabilitymeasurementtobeadifficultproblem.Thisthesissuggestsanavailabilitymeasurementmethod,whichsimplifiestheenvironment,andproposesthedefinitionofavailabilityandcomputingmethodbasedonupdatingfailureandnodefailure,thenillustrateswithexamples.Byanalyzingtheconceptandcharacteristicsofthecloudresourcemanagementandintegratingavarietyofresourceproperties,thisthesisresearchesintoresourceavailabilityinthecloudenvironment,andproposesaresourcepredictionmechanismandanavailabilitymethodbyintroducingqueuingtheoryandMarkovmodel.Theworkcontainstwoparts:thefirstpartpredictstheresourceavailabilityinthefuture,guidingresourceschedulingtoimproveitsefficiencyandresourceutilization;thesecondpartanalyzestheimpactthatsystemparametersproduceonresourceavailability,andprovidesreferencefordatamanagementstrategyforoptimizationofthesystemQoS.Keywords:Cloudcomputing,ResourceManagement,Resourceavailability,Resourceprediction,Updatingfailure,NodefailureVII1-1.................................................................................................................21-2Google............................................................41-3GFS..............................................................................................................81-4MapReduce..............................................................................................91-5WindowsAzure..........................................................................................112-1.......................................................................................................153-1....................................................................................263-2CloudSim...................................................................................323-3...................................................................................................323-4...................................................................................................334-1...........................................................................................354-2...............................................................................................384-3...........................................................................................414-4Markov.........................................................................................444-5Pn..........................................................................................................454-6....................................................................................461-1..............................................................