Research Of FAST TCP Network Congestion Control Protocol Based On Dynamic Model

With the continuous development of the internet, the emergence of new business, network capacity faces severe challenges. When available bandwidth is less than the demanded bandwidth, congestion will occur. The further network loads increase, the more serious congestion phenomenon becomes. Network congestion will degrade the performance of network, such as the increase of transmission delay and data packet loss rate, which has seriously influenced on business experience of the terminal customers. Research on network congestion to improve the quality of network has been becoming the current hotspot.The research on the network congestion protocols not only involves the stability, but also includes a series of dynamic behaviors, such as periodic oscillation, chaos, bifurcation, etc. In the time delay differential equations, periodic oscillation can cause Hopf bifurcation, which means system becomes unstable from steady state. Therefore, the study on the stability of internet congestion control system is of great significance.Based on the dynamic method, we have studied the equilibrium point of FAST TCP model, the stability of the limit cycle, Hopf bifurcation and the properties of bifurcation periodic solution. Firstly this paper introduces the definition of network congestion. Then the congestion control algorithm is described from two aspects consist of the source algorithm and routing algorithm. The principle of high-speed network congestion control mechanism is introduced in detail. We prove the existence of Hopf bifurcation in the mathematical model of FAST TCP by using dynamic analysis method, and selecting the communication latency as the bifurcation parameter to analyze the stability of the system. Then time delay control strategy to control bifurcation is taken to increase the critical value of communication time delay, so as to stabilize the signal transmission rate and routing buffer size, and effectively avoid network congestion within certain range of parameters. In this paper, the main innovation points are put as below:(1) The reason that the traditional TCP congestion control protocols are not applicable in high bandwidth delay product network is analyzed. In recent years, new protocols suitable for high-speed networks have been proposed, in which FAST TCP uses queue delay as congestion measure factor. Not only can it more accurately estimate the congestion, but also has high bandwidth utilization in maintaining the stability of the network.(2) The thesis makes an analysis on the dynamic model of FAST TCP congestion control system. With the application of the differential equation of knowledge and control theory, it is concluded that when the network time delay passes by a certain critical value, Hopf bifurcation occurs in the system, and a set of periodic solution bifurcating at the equilibrium point is obtained. From the practical point of view, this means that the steady signal transmission rate and routing buffer size are not guaranteed, which is not desirable. Bifurcation properties can be get by using the center manifold theorem and normal form theory. At last, the theoretical derivation is verified correctly according to the mathematics simulation.(3) The end-to-end congestion control in the network can be feasible. Adding a controller in a network model will not increase cost or network complexity from the point of view of technology. At the same time, it can significantly improve the stability of the network. So it has both theoretical significance and practical value. In order to eliminate the negative impact of the Hopf bifurcation for the system, time delay feedback algorithm is adopted in the paper. By selecting the suitable control parameters, the occurrence of Hopf bifurcation can be effectively postponed, therefore obtaining the ideal dynamics behavior.