Modeling the Performance of MapReduce Applications for the Cloud

Latin-American Journal of Computing

卷:2

Iván Carrera¹  Claudio Geyer² 

[1] Escuela Politécnica Nacional;

[2] Universidade Federal do Rio Grande do Sul;

关键词: MapReduce;    Cloud;    Hadoop;    FLOPs;    MRBS;    Performance;   

DOI  :  

来源: DOAJ

【 摘 要 】

In thelastyears,CloudComputinghasbecomeakeytechnologythatmadepossibletorunapplicationswithoutneeding todeployaphysicalinfrastructure.Thechallengewithdeploying distributedapplicationsinCloudComputingenvi-ronmentsisthatthevirtualmachineinfrastructureshouldbeplanned inatimeandcost-effectiveway.This workisasummaryofapreviousworkpresentedbytheauthors asaMaster’sthesis,withthegoalofshowingthattheexecution timeofadistributedMapReduceapplication,runningin aCloudcomputingenvironment,canbepredictedusingamathematical modelbasedontheoreticalspecifications.ThispredictionismadetohelptheusersoftheCloudComputingenvironmenttoplantheirdeployments,i.e.,quantifythenumberof virtualmachinesanditscharacteristics.Aftermeasuringtheapplication executiontimeandvaryingparametersstatedinthemathematical model,andafterthat,usingalinearregressiontechnique, thegoalisachievedfindingamodeloftheexecutiontime whichwasthenappliedtopredicttheexecutiontimeofMapReduce applications.Experimentswereconductedinseveralconfigurations andshowedaclearrelationwiththetheoreticalmodel, revealingthatthemodelisinfactabletopredicttheexecution timeofMapReduceapplications.Thedevelopedmodelis generic,meaningthatitusestheoreticalabstractionsforthecomputing capacityoftheenvironmentandthecomputingcostof theMapReduceapplication.

【 授权许可】

Unknown