Improving Grid computing performance prediction usingweighted templates

Understanding the performance behavior of Grid components to predict future Job submissions is considered one of the answers to automatically select computational resources to match users’ requirements maximizing its usability. Job characterization and similarity are key components in making a more accurate prediction. The purpose of this paper is to test how current data mining and statistical solutions that define jobs similarity perform in production Grid environments and to present a new method that defines template using two set of characteristics with different priorities and weights the templates prediction accuracy level for future use. The results show that the new method achieves in average a prediction errors than is 54 percent lower than those achieved by using dynamic templates.