云计算环境下基于.负载均衡的资源调度算法的研究与设计

《云计算环境下基于.负载均衡的资源调度算法的研究与设计》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、江苏大学硕士学位论文ABSRACTResourceschedulingisoneofthemostimportantresearchproblemsincloudcomputing．Thediversityofcloud—basedapplicationsandservers’resourcecapabilitiesleadtoloadimbalancebetweendifferentnodeswhichgreatlyimpactstheoverallperformanceofthecloudcomputingsyst

2、em．Thereforeproposingappropriateload·-balancingalgorithmstoachieveload··balancingbetweendifferentservershasbecomeanurgentproblem．Weproposetwoload-balancingalgorithmsaimingtoachievetwokindsofbad—balancingincloudcomputing，namelyphysicalmachine(PM)layerloadbalancin

3、gofresourcesandvirtualmachine(VM)layerload—balancingoftasks．Inordertoachievephysicallayerload-balancingofresources，weproposearesourceschedulingalgorithmbasedonlivevirtualmachinemigration．Inthisalgorithm,wehavestudiedthethreekeystepsinVMmigration：whentomigrate，wh

4、ichvirtualmachinetomigrate，wheretomigrateandproposedcorrespondingstrategiesforthem．Inordertodecidethemigrationtime，wepredicttheloadvariationtrendofaphysicalmachinebasedontrendofthecurrentloadandpredictitsfutureloadbydynamicallyadjustingtheincrementordecrementval

5、ue。Migrationwillbetriggeredifthecurrentloadandthepredictedloadarebothgreaterthanthemaximumthresholdorsmallerthantheminimumthreshold．InordertochoosetheproperVMtomigrate，weproposetotakebothCPUandmemoryresourceofphyrsicalmachinesandthevirtualmachinesrunningonthemin

6、toaccount．Besides，consideringthecostofmemorymigrationisquitelarge，wechoosetomigratethoseVIVIswithsmallmemoryconsumptiontoreducethemigrationcostaswellasmakethemostofphysicalresources．InordertofindthetargetPMtohostthemigratedVM，theresourcerequirementsofthemigrated

7、VMneedstobeconsidered．Moreover,weintroducetheprobabilitymechanismintotheselectionprocesstomakethesystemmoreresource-balanced．Furthermore，wesimulatedtheproposedalgorithmusingopenIII云计算环境下基于负载均衡的资源调度算法研究与设计sourceCloudSimsimulationplatformandsimulationresultsshowth

8、atouralgorithmCanachieveahighload-balancingdegree．Inordertoachievevirtualmachinelayerbad-balancing，weproposeadualweightedleast-connection(DWLC)schedulingalgorithm．The