Based On The Improved Stability Of FAST TCP Analysis And Research

With the rapid development of the network, people put forward new demands and challenges on the network. Applications of network become increasingly widespread, especially in recent years, online shopping is gradually increasing, the network also put forward new demands. At the same time, transmission of big data presents new challenges to network congestion. Some transmission of big data, the queuing delay will lead to imbalance, resulting in unfair. The predecessors proposed FAST TCP based on TCP Reno, and it is proved that having better convergence, balance, fairness and effectiveness.According to Steve Low described Fast TCP congestion control algorithms, this paper done a few following works:1.Based on the actual network environment, an window algorithms was improved. This algorithm is mainly on account of a delay in network transmission process each endpoint, consists of two parts, in the queue time, processing time delay into the queue, the queue delay processing.2.Gives a theoretical proof and analysis of the proposed algorithm. According FAST TCP model proved at an increased processing delay of the case, the algorithm convergence is still valid, but also calculated, under stable conditions, the parameters should meet conditions.3.Gives simulation and analysis of the improved algorithm, using NS2 simulation tool. This simulation is a continuous time model of a stable single pole single source network, in different delay environment, using NS2 to verify the improved window of the stability of the algorithm. Simulation results show that the theoretical analysis.In this paper, the main innovations are as follows:1.Proposed an improved version of the windows update formula, and proved the proposed algorithm;2.Simplified traditional windows update simulation model to a single pole single source network link model.According to the simulation results, in a large bandwidth high-latency environment, considering processing delay factor, according calculated parameters to set the parameters, can still guarantee the stable convergence of the algorithm.The significance of this study lies in the FAST TCP window control model improvements, taking into account the network processing delay factor for the future development of the network, especially in the control of the congestion window, provide some reference.