Self-managing systems able to dynamically re-configure withnrespect to time-varying workload mixes and changing system resourcenavailability are of growing importance in heterogeneous multicomputernenvironments. Advanced performance evaluation techniques are needed toninduce and assess the impact of such re-configurations wherenguaranteeing timeliness of reconfiguration activities is particularlynchallenging. A whole new class of methods supportingn“pro-active” adaptivity based on the predicted system statenat the re-configuration instant are needed to alleviate the shortcomingsnof “re-active” systems which bring reconfigurations inneffect after the system state has changed. In this paper we argue fornself-contained performance management of distributed applications, i.e.nintegrated performance tuning with the ability to automatically adaptnthe application behavior to the performance characteristics of thenexecution environment. Particularly, we study pro-active performancenmanagement for distributed simulation based on the Time Warp protocolnexecuting on a network of workstations (NOWs)
展开▼
