A Game-based Grid Resource Allocation Strategy

Grid is a significant technology of parallel distributed computing rise in recent years. Resource management is one of its key technologies. A good technology of resource management can use the resources on the grid more effectively, which can attract more resources owners to put their resources on the grid. The grid resource management is different from that in network environment, because the resources in grid have some features, heterogeneous, dynamic, autonomy and dichotomy. It is a hot problem that uses the economic theory in the grid resource management, and many researches had proved that it can do better. The math model in economic can be used to optimize and analyze the problem in resource management while lower the complexity of the problem.The thesis introduced the basic concepts in grid, research status domestic and abroad, and the three factors in grid technology. Based on the research status of grid resource management, the thesis analyzed the research using the economic model in resource management and the advantages of importing economic theory. As a part of mainstream economics, game theory is used to get the decision-making when groups of people interact and also the equilibrium problem, which fits the features of resources allocation in grid resource management.The contributions of the thesis are presented as follows. With regarding the situation that the grid consumer has budget restriction, formalize the problem that several consumers compete for one resource as a game with several players. From assuming the grid environment and set variables, the thesis founded a grid resource allocate mechanism and a biding policy model, proved the existent of Nash Equilibrium. Then gave a algorithm based on game theory. And in the end, the thesis did a performance simulation use the Gridsim toolkit.The results show that the grid resources allocate policy using game theory proposed in the thesis is feasible. It can improve the grid resources allocate performance compared with the other two algorithms.