A dynamic reservation based on scheduling method is proposed to reduce the disadvantage caused by inaccurate estimation on job run time. The method combines the reservation jobs and the computing resources in loose couple
and it enables the scheduler to recombine the reservation jobs and resources to reduce the idle resources caused by jobs completed in advance. Based on the dynamic reservation mode
the mind evolutionary computation(MEC)based algorithm is introduced into the scheduler. The algorithm sets the result obtained from the backfill algorithms as the initial population
and it can achieve better scheduling result by adopting assimilation and dissimilation operations. Simulations based on a real workload show that the proposed method is valid. Compared with the backfill algorithm
the proposed algorithm decreases both the average job wait time and the average slowdown by 68.5% and 66.9%
respectively.
关键词
Keywords
references
SRINIVASAN S, KETTIMUTHU R, SUBRAMANI V, et al. Selective reservation strategies for backfill job scheduling [C]∥8th International Workshop on Job Scheduling Strategies for Parallel Processing. Berlin, Germany: Springer-Verlag, 2002:55-71.
SHMUELI E, FEITELSON D. Backfilling with look ahead to optimize the performance of parallel job scheduling [C]∥9th International Workshop on Job Scheduling Strategies for Parallel Processing. Berlin, Germany: Springer-Verlag, 2003:228-251.
LAWSON B, SMIRNI E. Multiple-queue backfilling scheduling with priorities and reservations for parallel systems [C]∥8th International Workshop on Job Scheduling Strategies for Parallel Processing. Berlin, Germany: Springer-Verlag, 2002:72-87.
WARD W, MAHOOD C Jr, WEST J. Scheduling jobs on parallel systems using a relaxed backfill strategy [C]∥8th International Workshop on Job Scheduling Strategies for Parallel Processing. Berlin, Germany: Springer-Verlag, 2002: 88-102.
DROR F. Parallel workloads archive [EB/OL].[2007-09-20]. http:∥www.cs.huji.ac.il/labs/parallel/workload/index.html.
XUE Zhenghua,DONG Xiaoshe, MA Siyuan et al. An energy-efficient management mechanism for large-scale server clusters [C]∥Proceedings of the 2nd IEEE Asia-Pacific Services Computing. Piscataway, NJ, USA: IEEE, 2007:509-516.