预览加载中,请您耐心等待几秒...
1/2
2/2

在线预览结束,喜欢就下载吧,查找使用更方便

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

一种基于动态决策路径的网格任务调度算法 1.Introduction Withtherapiddevelopmentofbigdata,cloudcomputingandothertechnologies,moreandmoreapplicationsandservicesaredeployedoncloudplatforms,whichbringsgreatchallengestotheefficientallocationofresourcesoncloudplatforms.Asoneofthecoreissuesincloudcomputing,taskschedulinghasattractedmoreandmoreattentioninrecentyears.Theschedulingalgorithmplaysakeyroleintheefficiencyandperformanceofacloudplatform.Inthispaper,weproposeadynamicdecisionpath-basedgridtaskschedulingalgorithm. 2.RelatedWork Variousschedulingalgorithmshavebeenproposedinrecentyears,includingparalleltaskschedulingalgorithms,loadbalancingalgorithms,andoptimizationalgorithms.However,mostofthesealgorithmsonlyfocusonthestaticallocationoftasks,whileneglectingthedynamicchangesofthesystemstate.Someresearchershaveproposeddynamicschedulingalgorithms,suchasadaptiveschedulingandfeedbackscheduling.However,thesealgorithmsoftenhavehighoverheadandlackscalability. 3.DynamicDecisionPath-basedGridTaskSchedulingAlgorithm Toaddresstheaboveissues,weproposeadynamicdecisionpath-basedgridtaskschedulingalgorithm.Themainideaofthisalgorithmistodynamicallyadjustthedecision-makingprocessbasedonthechangingstateofthesystem.Specifically,thealgorithmconstructsadecisiontreebasedonthesystemstate,andusesthedecisiontreetodeterminetheschedulingstrategyforeachtask. 4.DecisionTreeConstruction Thedecisiontreeconstructionprocessincludesfoursteps:featureextraction,featureselection,decisionnodeselection,andpruning.Inthefirststep,weextractthefeaturesofthesystemstate,includingthenumberofnodes,theutilizationrateofCPU,memoryandnetworkbandwidth.Inthesecondstep,weuseinformationgaintoselectthemostimportantfeatures.Inthethirdstep,weselectthedecisionnodebasedontheselectedfeatures.Inthefourthstep,weusepruningtechniquestosimplifythedecisiontreeandavoidover-fitting. 5.TaskScheduling Oncethedecisiontreeisconstructed,theschedulingprocessisgivenasfollow: Firstly,whenanewtaskarrives,thesystemextractsthefeaturesofthecurrentstateandfindsthecorrespo