Title: Selective preemption strategies for parallel job scheduling

Authors: Rajkumar Kettimuthu, Vijay Subramani, Srividya Srinivasan, Thiagaraja Gopalsamy, D.K. Panda, P. Sadayappan

Addresses: Argonne National Laboratory, Argonne, IL 60439, USA; The University of Chicago, Chicago, IL 60615, USA. ' Microsoft Corporation, Redmond, WA 98052, USA. ' Microsoft Corporation, Redmond, WA 98052, USA. ' Altera Corporation, San Jose, CA 95134, USA. ' The Ohio State University, Columbus, OH 43210, USA. ' The Ohio State University, Columbus, OH 43210, USA

Abstract: Although theoretical results have been established regarding the utility of preemptive scheduling in reducing average job turnaround time, job suspension/restart is not much used in practice at supercomputer centres for parallel job scheduling. A number of questions remain unanswered regarding the practical utility of preemptive scheduling. We explore this issue through a simulation based study, using real job logs from supercomputer centres. We develop a tuneable selective suspension strategy and demonstrate its effectiveness. We also present new insights into the effect of preemptive scheduling on different job classes and deal with the impact of suspensions on worst case response time. Further, we analyse the performance of the proposed schemes under different load conditions.

Keywords: preemptive scheduling; parallel job scheduling; backfilling; high performance computing; supercomputers.

DOI: 10.1504/IJHPCN.2005.008032

International Journal of High Performance Computing and Networking, 2005 Vol.3 No.2/3, pp.122 - 152

Published online: 10 Nov 2005 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article