Gang scheduling with a queue for large jobs

192. B. B. Zhou and R. P. Brent, Gang scheduling with a queue for large jobs, Proc. 15th International Parallel and Distributed Processing Symposium, San Francisco, April 2001, 8 pp.

Paper: pdf (197K), ps (105K).


Applying gang scheduling can alleviate the blockade problem caused by exclusively space-sharing scheduling. To simply allow jobs to run simultaneously on the same processors as in conventional gang scheduling may introduce a large number of time slots into the system. In consequence the cost of context switches will be greatly increased, and each running job will only obtain a small portion of resources including memory space and processor utilisation, so no jobs will finish their computations quickly. Therefore, the number of jobs allowed to run in the system should be limited. In this paper we present some experimental results to show that, by limiting the number of real jobs timesharing the same processors and applying a backfilling technique, we can greatly reduce the average number of time slots in the system and significantly improve the performance of both small and large jobs.


Related papers are [169, 180, 181, 189, 194, 198, 202, 209].

Go to next publication

Return to Richard Brent's index page