SCHEDULING ALGORITHM FOR NONPREEMPTIVE MULTIPROCESSOR TASKS

This paper considers the problem of scheduling nonpreemptive multiprocessor tasks in a homogeneous system of processors. The problem proposed in this paper is different from the conventional scheduling problem, where each task requires only ''one'' processor whenever it is in processing. In our multiprocessor task scheduling problem, tasks can require more than one processor at a time for their processing, that is, tasks in our problem require simultaneously ''j'' arbitrary processors when they start to be processed, where 1 less-than-or-equal-to j less-than-or-equal-to k and k is a fixed number. Though multiprocessor tasks considered in this paper are assumed to be independent, i.e., the precedence relation does not exist among them. such a problem of scheduling nonpreemptive multiprocessor tasks is NP-complete. Therefore, a heuristic algorithm is investigated for this kind of problem and the worst performance bounds of the algorithm are also derived.