Distributed Computing Resource Scheduling Strategies Based On Multi-agent Theory For Large-scale Grid Systems

Grid is a novel Internet computing platform, which is emerged to the needs of problems sovling and applications. The objective of grid is to use the Internet to integrate all the online-computers from different places into a virtual super computer, so that all the information resource, computing resource, data resource , knowledge resource, storage resource, implement resource, expert resource and so on can be fully used by researchers and the ordinary people. Grid represents high technology and fundamentales tablishment. But for its complexity and completeness, it is still in the research stage and there are no criterions for grid so far. Grid has been the hotspot in the paralleler ocessing.In this paper, basing on the research of grid and multi-agent, analyzed the problems in grid resource scheduling by combined grid theory and multi-agent theory. And did some research in the resource scheduling for such system by coordination and negotiation.The main research in this paper can be classed as four parts:(1) Did a research on grid system from different aspects, including the origin and development, grid character, grid classes, grid architecture and grid technology.(2) Did some research on architecture and character of agent system and multi-agent system, and analyzed two hierarchical multi-agent architectures which were used widely. In addition, analyzed the logic architecture of the system which combined multi-agent technology with grid technology.(3) The fuzzy cognitive map was used to comstruct the coordination model for grid nodes to coordination their decisions, basing on this model, a coordination strategy was given. By using this coordination strategy, a resource scheduling algorithm based on multi-agent coordination for grid system was proposed. This algorithm can be distributed computed.(4) Basing on the decisions coordination strategy and the game theory, gave the algorithm of multilateral negotiation to form the task union for grid resource scheduling, and analyzed the validity. No upper level resource scheduling unit was needed in this algorithm. And with the distribution, it is suit for grid system.