The Study Of Multi-Agent Grid Resource Management Based On Virtual Supermarket

Grid computing is a new network computing platform after the appearance of World Wide Web. It provides a new infrastructure that can comprehensively share all kinds of resources. The resource in Grid has some characteristics: distributed, dynamic, heterogeneous, large-scale, multi-organization and so on, therefore Grid resource organization and management become the core issues of grid computing.Resource management is one of the key problems to provide high performance service for Grid. Compared with the conventional distributed system and network system, the Grid has different characteristics, and then Grid resource management has more complicated and more challengeable problems.Agent technology is an effective development method in distributed system. Taking the Agent technology into Grid technology is an effective way of solving grid problems.The start point of the paper is to resolve the current problems and defects in Grid resource management system. The paper presents Multi-Agent Grid Resource Management Architecture based on Virtual Supermarket, in which Grid resource is considered as commodity. The architecture combines with the business conception and management mechanism.Firstly, the paper refers to commodity marketing mode that is agency method step by step. On the basis of the idea of marketing mode, the whole Grid environment is carefully divided into many domains, including intra-area and inter-area; Secondly, combining with characteristics of virtual supermarket management model and mechanism, we give out a multi-Agent layered grid resource management model; Then, every functional Agent is described and designed referring to the theory of category optimization and shelf optimization in supermarket; Finally, a simulative experimentation based on GridSim, Which is a good simulation toolkit for grid resource scheduling, has been done to verify the efficiency. The simulation results show that the method using the learning ability of Agent can ceaselessly track the resource utilization service cost, so the method can optimize the resource selection.