A new DAG scheduling algorithm for heterogeneous platforms
Wahid Nasri, Wafa Nafti · 2012
Efficient task scheduling is one of the key factors for providing high performance on heterogeneous computing systems. However, due to many factors, such as the heterogeneity of such systems, solving efficiently such a problem is becoming a challenging task. In this paper, we address the problem of scheduling directed acyclic graphs (DAGs) on heterogeneous platforms composed of a set of processors grouped in a cluster of clusters (Multi-Clusters). Our main objective here is to develop a new DAG scheduling algorithm taking into account several characteristics of the target platform, including the heterogeneity of both computing powers of processors and communication links. Simulation results, based on both randomly generated graphs and platforms and real configuration based generated platforms, show that the proposed algorithm provides interesting performances and performs better than some well-known existing scheduling algorithms.