The Research And Realizating Of Network Congestion Control Based On Multi-Agent System

The network is more and more complex and heterogeneous, but the research of theory and methods for network management lags, due to the absence of analytic model. The monitor and control of high-speed network should be real-time and the traditional method is not applicable, so a method based on intelligent multi-Agent for network congestion control is presented in the paper. It takes full advantage of the distribution and cooperation of Agent to overcome the limitation of the centralized management and make the intelligent Agents to self-organize to control network congestion and guarantee the quality of service.The basic method of network management is introduced firstly and then its limitation and key problems are analyzed, and then the solution based on intelligent multi-Agent is brought forward through combining the theory of artificial intelligence and network management method.A network management system based on intelligent multi-Agent is designed and realized andit contains several aspects:1)function Agents: monitor Agent, control Agent, communication Agent, environment Agent2)communication language between Agent3)cooperation model for Agent4) knowledge baseThe system is applied to solve the deployment, performance, fault management for network. It is programmed in Java and can be operated on different network devices.,The topology discovery method is studied for network deployment management and an effect topology discovery algorithm is brought forward to help the other management work.The network congestion problem is investigated in the paper. The origin for congestion is analyzed and a congestion control method based on intelligent multi-Agent is presented to overcome the shortcomings of the traditional method. The congestion can be controlled fast and locally though the cooperation of local Agent and neighbours to avoid the deterioration. The efficiency is obvious especially in the network with large scale.A simulation experiment environment is set up to validate the congestion method. Through the analysis of experiment data for throughput, delay, packet-discarded rate, jitter, it can be proved that the performance is improved.