The Effect of Correlating Quantum Allocation and Job Size for Gang Scheduling

  • Authors:
  • Gaurav D. Ghare;Scott T. Leutenegger

  • Affiliations:
  • -;-

  • Venue:
  • IPPS/SPDP '99/JSSPP '99 Proceedings of the Job Scheduling Strategies for Parallel Processing
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

Gang scheduling is an effective scheduling policy for multiprocessing workloads with significant interprocess synchronization and is in common use in real installations. In this paper we show that signigicant improvement in the job slowdown metric can be achieved simply by allocating a different number of quanta to different rows (control groups) depending on the number of processes belonging to jobs in a given row. Specifically, we show that allocating the number of quanta inversely proportionally to the number of processes per job in that row results in 20 - 50% smaller slowdowns without significantly affecting mean job response time. Incorporating these suggestions in to real schedulers would require the addition of only a few lines of simple code, hence this work should have an immediate practical impact.