Accuratepredictionofserverloadisimportanttocloudsystemsforimprovingtheresourceutilization,reducingtheenergyconsumptionandguaranteeingthequalityofservice(QoS).Thispaperanalyzesthefeaturesofcloudserverloadandtheadvantagesanddisadvantagesoftypicalserverloadpredictionalgorithms,integratesthecloudmodel(CM)andtheMarkovchain(MC)togethertorealizeanewCM-MCalgorithm,andthenproposesanewserverloadpredictionalgorithmbasedonCM-MCforcloudsystems.Thealgorithmutilizesthehistoricaldatasampletrainingmethodofthecloudmodel,andutilizestheMarkovpredictiontheorytoobtainthemembershipdegreevector,basedonwhichtheweightedsumofthepredictedvaluesisusedforthecloudmodel.Theexperimentsshowthattheproposedpredictionalgorithmhashigherpredictionaccuracythanothertypicalserverloadpredictionalgorithms,especiallyifthedatahassignificantvolatility.TheproposedserverloadpredictionalgorithmbasedonCM-MCissuitableforcloudsystems,andcanhelptoreducetheenergyconsumptionofclouddatacenters.
