Modeling Scalable TCP friendliness to NewReno TCP

Summary Scalable TCP is a simple change to the traditional TCP congestion control algorithm which dramatically improves TCP performance in high speed wide area networks. The focus of this contribution is to model the performance of Scalable TCP algorithm sharing bottleneck link with a NewReno TCP. The analysis is performed during congestion avoidance phase, for both TCP flows, in steady state and near to the steady state. This model follows the existing model of NewReno TCP and is further expanded to predict the performance of congestion avoidance phase of the two TCP variants competing on a bottleneck link. The model is used in situations where congestion occurs due to buffer overflow (synchronous loss) and random loss (asynchronous loss). The applied model produced excellent information on how bottleneck link distributed over different TCP flavors. It predicts the Scalable TCP friendliness to NewReno TCP. Simulation measurements were carried out in order to verify the model accuracy. Long lived TCP flows traveling along high speed-wide area networks were used. The results of the applied model and the performed simulation measurements were in good agreement.