Based On Multi-agent Systems, Load Balancing And Fault-tolerant Management Research

With the development of Internet and communication, there has been a growing interest in the area of Agent and multi-agent system (MAS). When the MAS is more and more complex, the problem of the load balancing of the overall systems must be considered. This paper studies how to make the tasks assigning optimization in order to make the overall efficiency optimization in parallel computation system.Firstly, the paper researches the structure model of Agent and the technology of MAS step by step. Secondly, a architecture is created for the load balancing. The distributed information gathering tactics to collect load information is adopted, and the centralized decision tactics choose the destination nodes. Thirdly, because the characteristic of load based on Agent system is different from the multi-computer system, computation cost and communication cost must be considered at the same time in load balancing. The paper proposes the ideal of coupling degree to solve the conflict of those two costs, and the rationality of the ideal is proved by the experiments. Because it is a problem of combinatorial optimization, the simulated annealing algorithm is selected to solve the problem of load balancing. However, only approximately optimization result could get. So this paper proposes the optimized-result algorithm based on SA. And then, the advantage of the algorithm is proved by the experiment.Finally, this paper dose the preliminary research on the fault tolerance model of the system, establishes a redundant structure. Then, the corresponding treatment scheme according to the types of failure is given, and the failure is resumed.