Research On Congestion Control Algorithm In High-speed TCP/IP Networks

With computer networks being widely used in many fields, resources of the network are increasingly demanded by users that Internet traffic is constantly increasing at a high rate. Currently, network bandwidth capacity hinders transmission of business flows which network congestion has been main bottleneck against the development and application of network. Network congestion leads to reduction of performances including congestion window size, instantaneous queue length and response speed more easily. Consequently, network congestion control becomes research focus of communication field in recent years. Moreover, the approach to improving the network performance using control theory has been a frontier problem in international academic and engineering field.TCP/IP protocol is used by the data flows transferred 95 percent in the Internet, so the researcher at home and abroad proposed some excellent algorithms based on TCP/IP protocol, such as famous FAST TCP. Different from traditional TCP/IP protocol based on packet discarding, FAST TCP continuously monitors the time of sending packet and back ACK and predicts the condition of congestion according to the variation of transmission delay of network, and then regulates the sending rate of source that have more stability and faster transmission rate. However, while the distance existed in FAST TCP model is far away from the ideal equilibrium state, the variation of window show violent and cause the congestion of queue easily, the packet discarded, large variation in the queue delay, the unreasonable change of window function and oscillation phenomenon of congestion window. Aiming at these problems above, the study work and innovations are listed as follows:(1) The congestion control scheme named WTCP designed by a piecewise function is put forward based on the uncertainty of network window in FAST TCP. WTCP makes full use of the features of mathematics model and obtains a congestion control algorithm with steady-state response and transient response property. The stability of the algorithm is proved by basic principles and methods of control theory. Simulation results show that the performances including window size, queue length and their stabilities are better than those of similar algorithm under the bursty flow and without bursty flow transferred from the sources, respectively.(2) The effect of key parameters of network set to the performances is also researched. The congestion control algorithms parameters b and k set are proposed respectively. The values of b and k are analyzed and defined in the dissertation that the algorithms ensure the performances including the window size and queue length stable as well as response rate improvement. The stability of given algorithms are derived and proved by Routh-Hurwitz stability criterion and Lyapunov method. Simulation results show that the congestion control algorithms based on the key parameters set have improved the performances of network. At the same time, they provide the reference of parameters set for ensuring the network more stable. What's more, they have certain guiding significance for the configuration of router, switch and other important equipment of network in reality.(3) The asymptotic stability of FAST TCP with time domain model is researched. Considering that the variation of window is too violent and may cause the congestion of queue, even the packet is discarded and queue delay is too large, the variation of window function is unreasonable and may cause the vibration phenomenon when the distance existed in FAST TCP model is far away from the equilibrium state. The new congestion control algorithm is proposed based on above problems that the related parameter setting is derived to ensure the system asymptotic stability and the theoretical analysis and proof are provided by Lyapunov second method. Simulation results show that the improved algorithm makes the adjustment more stable and decreases the possibility of overflow and oscillating phenomenon greatly.(4) A congestion control Algorithm based on the network with big time lag is researched and developed. Based on the advantage of combination fuzzy control theory with Smith principle, Smith principle can compensate the influence of large time delay to stability of queue and fuzzy control calculates the packet loss probability dynamically, so congestion control algorithm with the combination fuzzy control with Smith principle is proposed. The stability of the algorithm is derived by Lyapunov method. Simulation results represent that the proposed algorithm improves the performance of network system, accelerates the response rate and decreases the high frequency vibration of queue length under different simulation scenarios.