Task mapping and priority assignment are critical to the efficiency of distributed multiprocessor systems. This paper presents a scheduling policy to perform these actions. The proposed method considers the stochastic rather than average character of the task execution times. The mapping problem is nondeterministic polynomial time (NP) hard. To keep the time to solution under control, a heuristic is used to map tasks and assign priority. The heuristic is based on the Tabu algorithm, and the search targets the minimization of a cost function including miss deviation. The proposed heuristic is evaluated for two case studies: a randomly generated benchmark, and a Global System for Mobile Communications (GSM) voice decoding.

The list of cited references consists of papers that are not recent. (Probably the reason for this is that in recent years the attention of the computer science (CS) community and the more recent publications on scheduling policies have moved from distributed to on-chip multiprocessor architectures.) My question for the authors is whether their approach can be applied to these architectures.