Task Scheduling Algorithm Using Optimal Duplication in Distributed Homogeneous Computing Systems

초록

Efficient DAG scheduling is crucial for achieving high performance in distributed homogeneous computing systems. Among the various task scheduling algorithms, full duplication scheduling algorithm considers the recursive duplication of the critical parents and reduces overall makespan. However, redundant duplications of parent nodes diminish idle time slots on processors and increase overall makespan. Therefore, in this paper, we propose a novel duplication based scheduling algorithm, called generalized duplication reduction algorithm (GDRA), to minimize redundant duplications, arisen in join nodes as well as fork nodes. In the experimental result, the proposed algorithm, GDRA, represents better performance under a small finite set of processors at high CCRs.