資源描述:
《基于模糊聚類的云任務調(diào)度優(yōu)化策略的研究》由會員上傳分享,免費在線閱讀,更多相關內(nèi)容在學術論文-天天文庫。
1、RESEARCHONOPTIMIZATIONOFCLOUDJOBSCHEDULINGSTRATEGYBASEDONFUZZYCLUSTERINGABSTRACTAsamajortechnicalsolutionofbigdataanddistributedcomputing,cloudcomputingprovidestheclusterresourcestotheendusersintheformofon-demandservices.Howtoallocateresourcesandschedulejobseff
2、ectivelyandreasonablyhavebecomethekeypointsanddifficulties.Itnotonlyhasaninfluenceontheefficiencyoncloudcomputingapplications,butalsohasadirectimpactontheoverallperformanceofcloudplatformandusersatisfaction.Inthecloudcomputingenvironment,clusternodesareheteroge
3、neous,diverse,uncertainandfuzzy,soitisdifficulttoaccuratelydescriberesourcesandjobs,anditleadstothecomplicationofon-demandresourceallocation.Resourceallocationandjobschedulingbasedonfuzzyclusteringhavebecomeahotresearchtopic.Astheclusterscaleisgettingenlarged,t
4、henumberofnodeshasalreadyreachedhundredsofthousands.Whendealingwithhigh-dimensionalandsuperhigh-dimensionalmatrixoperations,thetraditionalserialfuzzyclusteringalgorithmhastheproblemofhigh-amountcomputiation,loweffiency,computationspaceshortageandsoon,soit’sunab
5、letomeetthetimelinessrequirementsofresourceclusteringinthecloudcomputingenvironment.Therefore,optimizationandimprovementhavetobemadebasedontraditionalserialfuzzyclusteringalgorithm,inordertosuitcloudcomputingbetter.Firstly,studyonthetraditionalserialfuzzycluste
6、ringalgorithmintheIIIcloudenvironment,compareandanalyzethreetraditionalclusteringalgorithmsandsummarizetheirfeaturesandshortcomings.Anoptimizationandconcurrencystrategybasedonfuzzyequivalencerelationandfuzzysimilarityprocessing,itcanreducetheamountofcomputation
7、,improvecomputingefficiency,andshortenthetime-consuming.Experimentalresultsshow,comparedtotheexistingmethods,thestrategycaneffectivelysolvetheproblemofhighoverheadinthesmallandmedium-scalecloudresourceclusteringdivisionprocess.Secondly,inordertodealwiththeprobl
8、emofdifficultytodivideresourceclusteringcausedbyhigh-amountcomputationandmemoryshortageinthetraditionalfuzzyclusteringalgorithm,aconcurrencyfuzzyclusteringalgorithmbasedonMa