简介：Inacomputationalgrid,jobsmustadapttothedynamicallychangingheterogeneousenvironmentwithanobjectiveofmaintainingthequalityofservice.Inordertoenableadaptiveexecutionofmultiplejobsrunningconcurrentlyinacomputationalgrid,weproposeanintegratedperformance-basedresourcemanagementframeworkthatissupportedbyamulti-agentsystem(MAS).Themulti-agentsysteminitiallyallocatesthejobsontodifferentresourceprovidersbasedonaresourceselectionalgorithm.Later,duringruntime,ifperformanceofanyjobdegradesorqualityofservicecannotbemaintainedforsomereason(resourcefailureoroverloading),themulti-agentsystemassiststhejobtoadapttothesystem.Thispaperfocusesonapartofourframeworkinwhichadaptiveexecutionfacilityissupported.Adaptiveexecutionfacilityisavailedbyreallocationandlocaltuningofjobs.Mobile,aswellasstaticagentsareemployedforthispurpose.Thepaperprovidesasummaryofthedesignandimplementationanddemonstratestheefficiencyoftheframeworkbyconductingexperimentsonalocalgridtestbed.