Performance evaluation of heuristics for scheduling pipelined multiprocessor tasks

Abstract

This paper presents the evaluation of the solution quality of heuristic algorithms developed for scheduling multiprocessor tasks in a class of multiprocessor architecture designed for real-time operations. MIMD parallelism and multiprogramming support are the two main characteristics of multiprocessor architecture considered. The solution methodology includes different techniques including simulated annealing, tabu search, as well as well-known simple priority rule based heuristics. The results obtained by these different techniques are analyzed for different number of jobs and machine configurations.