Task scheduling algorithm with minimal redundant duplications in homogeneous multiprocessor system

Abstract

The performance of a homogeneous multiprocessor system depends on an algorithm which schedules input DAG. Among various task scheduling algorithms, ones that are using full duplication provide superior performance by duplicating tasks recursively. However, redundant duplications lead to increase of total execution time and decrease of processor idle time. In this paper, we propose GDRA which reduces total execution time and increases processor idle time by eliminating redundant duplications. Comparisons with other full duplication algorithms, such as CPFD, SD, and MDJ, show that GDRA provides the better performance as the CCR increases. © 2011 Springer-Verlag.