Resource Evaluation And Selection Strategy Analysis And Realization In Grid Workflow

Grid technology allows resource sharing among several entities, but one of main problems of grid workflow is selecting the best resource to run a specific job. The challenges for the best resource selection involves analysis of several factors such as prediction time to run a job, access restriction to resources, and cost to use resources etc. In addition, resource selection must consider uncertainties associated with each resource and answer questions related to resource reliability, prediction error probability, and cost error probability to access a resource. The resources scheduling problem of task scheduling of grid workflow has been tackled in literature in a number of different ways. Nevertheless, each approach solves this problem just considering only or two of the factors that can lead to the optimal selection, don't completely consider those important factors as mention before, so they can't satisfy the user's multi-dimensional QoS requirements. Considering that the grid workflow resource scheduling problems belong to multi-objective decision-making, thus this thesis put forward three multi-dimensional QoS guided resource evaluation and selection algorithm that used for single task scheduling of grid workflow through discussing, analyzing and modifying the method of solving kind of MADM question, namely RESA that emphasize particularly on object decision-making, RESA-E that is completely subject decision-making, RESA-LE that emphasize particularly on subject-making. They all ensure that they would search the optimal resource which satisfy the user's requirements, not only satisfying the cost, time optimal, but also satisfying user's personal special QoS requirement, finally, carry them out with Matlab programs and draw an conclusion that RESA is the optimal algorithm used for solving question of grid workflow resource evaluation and selection through comparing with others. RESA provides a theoretical basis and application prospects for automated tools for resources selection and solution of resources diversity of grid.