摘要
Multipleperformancerequirementsneedtobeguaranteedinsomereal-timeapplicationssuchasmultimediadataprocessingandreal-timesignalprocessinginadditiontotimingconstraints.Unfortunately,mostconventionalschedulingalgorithmsonlytakeoneortwodimensionsofthemintoaccount.Motivatedbythisfact,thispaperinvestigatestheproblemofprovidingmultipleperformanceguaranteesincludingtimeliness,QoS,throughput,QoSfairnessandloadbalancingforasetofindependenttasksbydynamicscheduling.Webuildaschedulermodelthatcanbeusedformulti-dimensionalscheduling.Basedontheschedulermodel,weproposeaheuristicmulti-dimensionalschedulingstrategy,MDSS,consistingofthreesteps.Thefirststepcanbeofanyexistingreal-timeschedulingalgorithmthatdeterminestoacceptorrejectatask.Instep2,weputforwardanovelalgorithmMQFQtoenhancetheQoSlevelsofacceptedtasks,andtomakethesetaskshavefairQoSlevelsatthesametime.AnothernewalgorithmITLBisproposedandusedinstep3.TheITLBalgorithmiscapableofbalancingloadandimprovingthroughputofthesystem.ToevaluatetheperformanceofMDSS,weperformextensivesimulationexperimentstocompareMDSSstrategywithMDSRstrategy,DASAPandDALAPalgorithms.ExperimentalresultsshowthatMDSSsignificantlyoutperformsMDSR,DASAPandDALAP.
出版日期
2009年03月13日（中国期刊网平台首次上网日期，不代表论文的发表时间）
