Dynamic Sequencing of A Multi-Processor System: A Genetic Algorithm Approach

Abstract

The dynamic sequencing of jobs through a multiprocessor system is one to which little attention has been paid, in contrast to the extensive literature on static sequencing problems. Yet in many real problems, to assume, as the static model does, that we know about all jobs that will arrive in the course of a processing cycle is hardly realistic.

Existing solutions usually assume a queueingtheoretic orientation, rather than an optimization one, in which the decision as to which job should be processed is made on the basis of some simple selection criteria, such as First-Come First-Served, or Shortest Processing Time.

Here we investigate the use of a Genetic Algorithm (GA) to solve the successive sequencing problems generated by finding a near-optimal sequence for those jobs available just before successive event times—the times at which the job being processed on the first processor completes its processing. Some comparisons are made between using the GA approach versus some simple rules.

On sabbatical leave in Dept. of Computer Science, City University, London, for the period of this research.