Predicting grid performance based on novel reduct algorithm

Abstract

Because of the irregular characteristic of Grid environment, we are unable to predict the performance using the traditional method. In this paper, we propose a novel method for predicting the performance in Grid Computing environment. The method, based on frequencies of application attributes appeared in discernibility matrix collected during a period of time; predict the applications performance that the traditional methods can't obtain. We use the novel method in Data Ming Grid and obtain better result than traditional methods. The results of the experiment show that the use of reduct algorithm can process uncertain problem in Data Mining Grid. The theoretical foundation of ruduct provides a feasible solution to the problem of predicting Data Mining Grid. © 2008 Springer-Verlag Berlin Heidelberg.