Mining Optimal Resource Combination in Computational Grid

Resource allocation is the key technology in computational grid. How to effectively allocate and schedule the grid resources is a challenge for a grid computing system. To the grid service providers, two issues should be addressed to maximize their profits: 1) how to choose the optimal resource combination; 2) how to adjust the resource price under certain resource combination. While most of the existing grid literatures only consider the second problem, this paper focuses on the first one. This paper provides a novel approach to optimize the resource combination in computational grid. By analyzing historical grid trade data, the proposed method can mine the important associations between different kinds of grid resources based on principal component analysis (PCA). Another contribution of this paper is that it attempts to measure the fitness of certain grid resource combination. Moreover, this paper shows how to adjust the grid resource combination when its fitness is not good enough. Simulation results prove the well performance of the proposed method