Self-adaptive Task Allocation and Scheduling of Meta-tasks in Non-dedicated Heterogeneous Computing
Authors: M. Wu, X.-H. Sun
Date: January, 2004
Venue: International Journal of High Performance Computing and Networking , vol. 2, no. 2/3/4, pp. 186-197
Type: Journal
Abstract
The efforts to construct a national scale Grid computing environment have brought unprecedented computing capacity and complicacy. Exploiting this complex infrastructure requires efficient middleware to support the execution of distributed applications, which presents the challenge on how to schedule tasks in shared heterogeneous systems. Most existing scheduling systems are based on pre-determined estimation of task completion time and resources availability. They may not provide appropriate scheduling if the underlying computing resources present an abnormal usage pattern during an application execution. For solving long-running applications in a large-scale Grid environment, abnormal usage of some resource may not be uncommon. We have proposed the development of the Grid Harvest Service (GHS) performance evaluation and task scheduling system in our previous work. In this study, we present a novel dynamic self-adaptive scheduling algorithm and its implementation under GHS. Scheduling and re-scheduling algorithms and mechanisms are carefully investigated. Experimental results show that, equipped with these new scheduling mechanisms, GHS outperforms existing systems considerably in scheduling large applications in a non- dedicated heterogeneous environment.