Research Of Congestion Control System In High Speed Networks Based On Dynamic Model

As internet developing rapidly, network scale expanding, types of application further enriching and the number of users increasing sharply, network congestion is a growing problem. Due to network congestion directly leads to poor network performance, network congestion control is the main method to improve it. Meanwhile, traditional network congestion control algorithm can seriously degrade performance in high-speed network, because the development of next-generation network and the growing requirement of data transmission in various new applications. Therefore, researching congestion control algorithms that can be applied to high-speed network has been a new hotspot.First, using nonlinear dynamics analysis means, the paper built a nonlinear dynamics model of high-speed network congestion control system. Then complex nonlinear dynamics behaviors of the system were studied and controlled by means of nonlinear dynamics and control theories. Comparing with traditional network congestion control researching method based on measurement and simulation, more precise conclusion and internal mechanism in the network congestion control system can be reached with the model. With it, essential reasons of high-speed network congestion were explained and corresponding algorithms can be designed, so it has theoretic importance and application value. Main work and innovations of this paper are shown as follows:(1) High-speed network congestion control protocol TCP LogWestwood＋has been further studied. The paper analyzes logarithmic increase strategy of congestion window (cwnd) and fluid-flow model in TCP LogWestwood＋. We have researched the overall network congestion closed loop feedback control mechanism in detail when TCP LogWestwood＋is used in the source and Random Early Detection (RED) is adopted in the link router. Then this paper develops a discrete time dynamic feedback model for a simple dumbbell-shaped topology network with TCP LogWestwood＋/RED.(2) In the network congestion control system with TCP LogWestwood+/RED, parameters stability intervals and complex nonlinear dynamic behaviors have been studied by nonlinear dynamics theories, such as bifurcation and chaos. Choosing RED control parameters, number of connections, propagation delay as bifurcation parameter, bifurcation diagram and Lyapunov exponent are used to directly demonstrate periodic doubling bifurcation and chaos.(3) In order to eliminate the adverse effect of chaotic behaviors in network congestion control system with TCP LogWestwood+/RED and stabilize chaotic orbits, the paper introduces a hybrid control algorithm with parameter perturbation and state feedback. Theories analysis and numerical simulation have shown that this method can delay bifurcations effectively, and even eliminate chaotic behaviors perfectly with parameters limit, then stability intervals are widened.