An improved preference method for the FAST TCP protocol parameter

Aiming at the weaknesses of FAST TCP in choosing protocol parameter for static mapping. A slow time-scale iterative search preference parameter method based on the difference between each path traffic service intensity and its transmission bandwidth is proposed. The stochastic model and its fluid model was derived, in which each path TCP connection followed Poisson process and the size of transport files obeyed the exponential distribution. A Lyapunov function including TCP connection’s mean waiting time was constructed to prove that, for those links which are in heavy traffic, if each path’s traffic service intensity was equal to its transmission bandwidth, the number of each path’s active TCP connection would stay stable, and the average waiting time will reach the minimum. Thereby the proposed method is theoretically feasible. Moreover, NS-2 simulations are presented to verify this effective method.