The evaluation of parallel job schedulers hinges on two things: the use of appropriate metrics, and the use of appropriate workloads on which the scheduler can operate. We argue that the focus should be on on-line open systems, and propose that a standard workload should be used as a benchmark for schedulers. This benchmark will specify distributions of parallelism and runtime, as found by analyzing accounting traces, and also internal structures that create different speedup and synchronization characteristics. As for metrics, we present some problems with slowdown and bounded slowdown that have been proposed recently.
|Original language||American English|
|Title of host publication||Job Scheduling Strategies for Parallel Processing - IPPS/SPDP 1998 Workshop, Proceedings|
|Editors||Dror G. Feitelson, Larry Rudolph|
|Number of pages||24|
|ISBN (Print)||3540648259, 9783540648253|
|State||Published - 1998|
|Event||4th Workshop on Job Scheduling Strategies for Parallel Processing held in conjunction with IPPS/SPDP 1998 - Orlando, United States|
Duration: 30 Mar 1998 → 30 Mar 1998
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||4th Workshop on Job Scheduling Strategies for Parallel Processing held in conjunction with IPPS/SPDP 1998|
|Period||30/03/98 → 30/03/98|
Bibliographical notePublisher Copyright:
© Springer-Verlag Berlin Heidelberg 1998.