An Effective Task Scheduling Method with Link Contention Constraints for Heterogeneous System

Abstract

Sufficient and various computing resources make heterogeneous system suitable for parallel and distributed applications. A task scheduling problem is to arrange tasks of application on computing resources and aim to achieve minimum schedule length. Many effective scheduling algorithms have been proposed, but most of them assume that the network is fully-connected and contention free. In order to make this problem more practical, we take the link contention constraints into consideration. In this paper, we propose an effective and efficient Communication Look-ahead Scheduling (CLS) algorithm extended from list-scheduling algorithm. CLS contains two scheduling phases. In the first phase, we prioritize tasks with a new value which integrates both information from the concept of critical path and the communication behavior of each task. In the second phase, we propose a processor selecting mechanism with communication look-ahead manner to allocate tasks. According to our performance evaluations, CLS is superior to other algorithms both in effectiveness and efficiency.