A simulated annealing approach to parallel processor scheduling problems with precedence relations

Abstract

Abstract The purpose of this paper is to describe a new heuristic method based on the simulated annealing (SA) approach, to solve parallel processor scheduling problems. Jobs to be scheduled on parallel identical processors have dependence relationships and the goal is to minimize the makespan. We studied various parameter settings, such as the initial schedule, perturbation scheme and annealing schedule in SA approach, and their effect on the performance measure (i.e., the makespan). Also, we compared the SA approach to the traditional heuristic methods such as the longest processing time first (LPT), Hu's level scheduling (LS), and the critical path first (CP) procedures. Experimental results show that the SA approach performs much better than these heuristic approaches, and provides very competitive solutions when compared to optimal solutions generated from the Branch and Bound approach.