An Efficient Resource Selection Strategy Based on Load Forecasting in Computational Grid

摘要

Grid computing provides a virtual framework for controlled sharing of resources in a virtual organization. In a dynamic grid environment, the real challenge is to efficiently manage the application execution on distributed resources by appropriate resource selection and allocation strategies. We present a novel resource selection strategy by analyzing the load history of trusted resources and forecasting the future load of the resources in a virtual organization. The double exponential smoothing approach is employed for load forecasting on selected trusted resources. The prediction accuracy is achieved by a load adaptive factor that adjusts itself with dynamism according to the changes in the pattern of the load on the resource. The proposed method employs prediction to determine the availability of resources in the near future and allocates the best suitable resource that satisfies the job requirement thus eliminating run time failures and unnecessary job migration. Simulation results show that the resource selection strategy improves the success rate of the jobs submitted to the grid by utilizing the resources to its maximum potential. The load forecasts are reliable for short term predictions and achieve a good prediction accuracy of about 84% even under high load uncertainty conditions.