Research On Congestion Control Strategy For Complex Network

With the increasing demand from the community for the information, the Internet is showing explosive growth, making the traffic grows rapidly, thus, the congestion problem of the Internet becomes one of the major barriers in using Internet. The traditional TCP congestion control strategy is an end-to-end control strategy, the transmitters detect congestion basing on the arrivals of duplicate ACK or signals from retransmission timeout, and the nodes discard packets when they overflowed the buffers to ease congestion. The response to the congestion is slow and will easily result in severe congestion. Therefore, we need new congestion control strategys to solve this problem.This paper begins with describing scale-free features, clustering properties and small-world properties of the complex network, then makes a brief introduction of various statistical parameters of the complex network, and gives an overview of related congestion control strategies of the Internet. On this basis, this paper proposes complex network theory based strategies for congestion control on the Internet (CSCI).CSCI is including substances at four aspects: congestion detection, congestion avoidance, congestion mitigation and fairness-based bandwidth allocation. In the past congestion control strategy, only the sender and the receiver are allowed to respond to the congestion, the response to the congestion is hysteretic. To resovle this problem, this paper proposes the cache queue based congestion detection and RTT-based congestion detection, so that intermediate nodes are also involved in the congestion control system, so they can greatly reduce the time to detect congestion. In consideration on how to avoid congestion, this paper put forward four kinds of value assignment strategies: the shortest path, the clustering coefficient, the betweenness and the greedy weight function. So the system can choose the node with the suitable value as the next node for a data in transmission to step in, thus there is a variety of ways to avoid the congestion.According to the principle of easing congestion rapidly when it occurs, this paper presents user feedback based strategy and priority delivery based strategy. They have higher efficiency to relieve the congestion. To resovle the problem that small data streams are "drowned" by large data flows in the congestion nodes, this paper presents fairness based bandwidth allocation algorithm, which can improve the fairness of bandwidth allocation after congestion occurs. Finally, on the Matlab platform, the proposed congestion control strategies are simulated. The simulation results show that, CSCI has higher throughput and lower average latency and is able to give real-time and effective control to congestion in the network.