A Dynamic Multi-Strategies Scheduling Algorithm for Heterogeneous Computing Systems

Efficient scheduling algorithms play an essential part in heterogeneous computing systems (HeCS) to achieve high performance. So we proposed a new efficient dynamic multi-strategies scheduling algorithm (DMS). The basic idea of this algorithm is to reduce communication in the priority calculation phase and task mapping phase. In the priority calculation phase, the algorithm increases the weight of the task. During the task mapping phase, three approaches for mapping tasks are proposed, and the method which reduces the most communication time among the three approaches is chosen as the task mapping standard. The results of the algorithm presented in this paper demonstrate improved performance in terms of speedup, and schedule length ratio.