An online and predictive method for grid scheduling based on data mining and rough set

Abstract

Since Grid is a distributed and heterogeneous environment, scheduling and resource management are important in Grid. One of the fundamental problems in Grid is designing a suitable method for management of resources. Many approaches have been proposed to increase performance of scheduling process, but each method has special conditions and they act well only in some special conditions. Moreover for resources scheduling, most of them use GIS's data that maybe is encountered with old data. In this paper, we use an online approach for finding resources with less time spending for resource discovery rather than other proposed methods; and then by using Rough Set and Decision Tree data mining technique, in order to classification of Grid nodes, scheduler will select proper nodes for desired job based on job's condition. This approach also has a fair treat in dealing with The Least Deadline jobs. The obtained results show this approach is one of the promising methods for resource selecting in scheduling phase with high accuracy and performance. © 2009 Springer Berlin Heidelberg.