Research Of Multi Agent-based Dynamic Load Balancing Approach And Application

As the development of information technology and mobile communication technology, computer technology has developed to the network-centric information exchange age. Network service sites are no longer confined to provide the traditional type of services, such as HTTP, FTP, new type of services such as streaming audio and video, online games, also can be seen anywhere. With the diversification of network services, massive of information and real time transformation of user data exchange, service providers need to deal with the problem that how to meet the growing user request.However, with the advance of information technology, the key to restricting network Qos is no longer the bottleneck of network communication, but the server performance. The traditional single-server system has a very limited ability to handle user requests, and to purchase large servers will make the original servers idle, resulting in a waste of resources.Based on the above requirements, building high performance and high availability server play the key role to the solution of this problem. Cluster can make full use of existing servers to achieve high performance computing systems. Furthermore, it has a series of advantages, such as autonomy, parallelism, scalability, and transparency, and it has become an efficient parallel/distributed computing resources and is widely welcomed. Now the key issues which need to be solved are how to make full use of cluster resources and how to shorten the response time of user requests. Load balancing technology is an effective way to solve this problem.In this paper, based on researching clusters, load balancing technology and agent technology, using transfer policy, selection policy, location policy, information policy, etc,a dynamic load balancing algorithm based on multi-agent is designed, applying centralized scheduling policy and regarding JADE platform as a test bed, combining heuristic scheduling strategy and active scheduling policy, request analysis agent, load-info gather agent, load-info monitor agent, load transfer mobile agent and parameter adjustment mobile agent are developed to complete load balancing of system, and verify the effectiveness of the algorithm.In addition, we study the dynamic adjustment of parameters, session data support, and fault-tolerant mechanism of MAS preliminary. In order to achieve load balancing of system efficiently, using fuzzy control technology to adjust the parameters.