Enhanced Adaptive Scheduling for the Grid Harvest Service

Abstract

Grid technology and applications become mainstream distributed computing research in recent years. The Grid Harvest Service, GHS, is one of non-dedicated computing grid scheduling systems; it has been widely used in Grid research field. The contribution of GHS is providing appropriate prediction for long-term applications different from AppLes which is designed for short-term predictions. GHS maps metatasks using the Min-Min algorithm in a uniform log-term application. Min-Min is a highly efficient algorithm in an uniform workload environment that leads to load unbalance and low performance when the workload is not uniform. In order to solve the problem, a novel task scheduling algorithm is proposed in this work, which is an adaptive task scheduling algorithm based on Min-Min and Max-Min (A-MM). A-MM merges the high efficiency of traditional Min-Min scheduling with load balance of traditional Max-Min scheduling. The simulation and experimental result show that A-MM has better performance and scalability than the Min-Min of GHS.