Task Execution Availability Prediction in the Enterprise Desktop Grid

Abstract

Desktop grid uses an idle cycle of desktop PC’s in Intranet and Internet environments for large-scale computation. Heterogeneity and volatility of hosts within this environment are lead to consider issues of availability of resources. Resource availability is critical for the reliability and responsiveness of services. In this paper, two prediction systems have been introduced for task execution availability of resources in the enterprise desktop grid platform. The first one is based on cellular automata and the other one according to Bayesian network. The accuracy of proposed prediction systems is evaluated via four real desktop grids. In spite of highly volatile environment, the meaningful and robust prediction results have been gained. A comparison between two prediction systems indicates that cellular automata have a better behaviour than Bayesian network. It can predict the behaviour of resources in the desktop grid with the prediction accuracy of %95.5.