Congestion Control Strategies Based On Minority Game With Local Information

With the rapid development of Internet and emergence of network applications, network congestion problem is increasingly outstanding. So the strategies of congestion control are very important for current TCP/IP network.Although traditional TCP Congestion control mechanism has been playing an important role in preventing network congestion, it also exposes much intrinsic limitation, like feedback delay, rough granular congestion control and low resource utilization etc. To solve these problems, many congestion control algorithms is proposed in recent years, but some of these algorithm still can't give a satisfied solution for un-fairness of resource allocation in multi-bottleneck network and low link utilization.From a perspective of resource allocation, TCP congestion control is a mechanism to allocate resource to each sender in the network reasonably. Each sender's choice of transmit speed will depend on those of other senders, and the network congestion status is determined by all the choice together. So it is necessary to address the behavior of senders in the framework of game theory.Considering the local information of network, a congestion prediction model based on minority game (MG) with local information is proposed. To feedback the Congestion condition in time and clearly, an exact congestion feedback mechanism is designed. Considering the link utilization and unfairness, a congestion control algorithm based on proposed model is designed. The result of simulation indicates that the proposed algorithm has good performance on link utilization and fairness.