TCP fairness enhancement through a mathematical parametric model

For a particular Internet connection the number of outstanding packets is controlled by TCP´s current window size. During congestion avoidance phase, the congestion window is increased by 1 packet roughly during every RTT. It means that in each second, the throughput of a typical node can be increased by 1/RTT pkts. Therefore, the sources with smaller RTT experience higher throughput than those with long RTT. Using Opnet Modeler simulator we show that how this RTT differences between competing traffic streams, can be the dominant factor in relative throughput. In order to decrease this discrimination, we introduce a parametric algorithm for congestion window (cwnd) and therefore for rate control. We show that using an appropriate formulation derived from overall optimization problem, the network implicit objective function provides a Lyapunov function for the dynamic system defined by the rate control algorithm. Then we show that how this mathematical model are being used to address fairness problem of the rate control algorithms in the Internet. Our simulation results show the success of the proposed algorithm.