JSSPP 2002: Job Scheduling Strategies for Parallel Processing pp 24-37 | Cite as
Preemption Based Backfill
Abstract
Recent advances in DNA analysis, global climate modeling and computational fluid dynamics have increased the demand for supercomputing resources. Through increasing the efficiency and throughput of existing supercomputing centers, additional computational power can be provided for these applications. Backfill has been shown to increase the efficiency of supercomputer schedulers for large, homogenous machines[1]. Utilizations can still be as low as 60% for machines with heterogeneous resources and strict administrative requirements. Preemption based backfill allows the scheduler to be more aggressive in filling up the schedule for a supercomputer[2]. Utilization can be increased and administrative requirements relaxed if it is possible to preempt a running job to allow a higher priority task to run.
Keywords
Expansion Factor Priority Task High Priority Task Queue Time Lower Priority TaskPreview
Unable to display preview. Download preview PDF.
References
- [1]D.G. Feitelson and L. Rudolph. Parallel jobs cheduling: Issues and approaches. Lecture Notes in Computer Science: Job Scheduling Strategies for Parallel Processing, 949, 1995. 24Google Scholar
- [2]G. Berry. Preemption in concurrent systems. FSTTCS’ 93, Lecture Notes in Computer Science 761, pages 72–93, 1993. 24Google Scholar
- [3]D.G. Feitelson and M. Jette. Improved utilization and responsiveness with gang scheduling. Proceedings of the IPPS’ 97 Workshop on Job Scheduling Strategies for Parallel Processing, 1997. 26Google Scholar
- [4]Uwe Schwiegelshohn. Preemptive weighted completion time scheduling of parallel jobs. Lecture Notes in Computer Science: 4th Europeon Symposium on Algorithms, 1136:39–51, 1996. 28Google Scholar
- [5]Anat Batat and Dror G. Feitelson. Gang scheduling with memory considerations. In 14th International Parallel and Distributed Processing Symposium, 2000. 28Google Scholar
- [6]D. Jackson. The Maui Scheduler. Technical report. http://supercluster.org/projects/maui. 29, 31
- [7]D. Jackson, Q. Snell, and M. Clement. Core algorithms of the maui scheduler. Lecture Notes in Computer Science: Job Scheduling Strategies for Parallel Processing, 2221:87–102, 2001. 29CrossRefGoogle Scholar
- [8]Q. Snell, M. Clement, D. Jackson, and C. Gregory. The performance impact of advance reservation metascheduling. Lecture Notes in Computer Science: Job Scheduling Strategiew for Parallel Processing, 1911, 2000. 29CrossRefGoogle Scholar