摘要
Accuratepredictionofserverloadisimportanttocloudsystemsforimprovingtheresourceutilization,reducingtheenergyconsumptionandguaranteeingthequalityofservice(QoS).Thispaperanalyzesthefeaturesofcloudserverloadandtheadvantagesanddisadvantagesoftypicalserverloadpredictionalgorithms,integratesthecloudmodel(CM)andtheMarkovchain(MC)togethertorealizeanewCM-MCalgorithm,andthenproposesanewserverloadpredictionalgorithmbasedonCM-MCforcloudsystems.Thealgorithmutilizesthehistoricaldatasampletrainingmethodofthecloudmodel,andutilizestheMarkovpredictiontheorytoobtainthemembershipdegreevector,basedonwhichtheweightedsumofthepredictedvaluesisusedforthecloudmodel.Theexperimentsshowthattheproposedpredictionalgorithmhashigherpredictionaccuracythanothertypicalserverloadpredictionalgorithms,especiallyifthedatahassignificantvolatility.TheproposedserverloadpredictionalgorithmbasedonCM-MCissuitableforcloudsystems,andcanhelptoreducetheenergyconsumptionofclouddatacenters.
出版日期
2018年05月15日(中国期刊网平台首次上网日期,不代表论文的发表时间)