A novel task scheduling algorithm for distributed heterogeneous computing systems

Abstract

This paper proposes a novel task scheduling algorithm to exploit the potential of parallel processing, allowing for system heterogeneity in distributed heterogeneous computing environments. Its goal is to achieve maximizing parallelization and minimizing communication. Due to that the algorithm avoids from the max-min anomaly in the parallelization problem and exploits schedule holes, it could produce better schedules than those obtained by existing algorithms. Experimental results are presented to verify the preceding claims. Three comparative algorithms are applied to demonstrate the proposed algorithm's effectiveness. As the system's heterogeneity increases, the performance improvement of the proposed algorithm becomes more outstanding than that of others. Therefore, the proposed. scheduling algorithm may be used in designing efficient parallel environments for those situations where the system heterogeneity is the system performance bottleneck. © Springer-Verlag Berlin Heidelberg 2006.