Modeling of TCP-Rab throughput and its empirical study

In our previous work, a new TCP protocol which is named receiver advertisement based TCP (TCP-Rab) is proposed. The design and implementation of Rab's algorithm are described in details in the literature [1]. In this paper, we made only a little modification on TCP-Rab, the pivot issue of this paper is focused on the throughput model and empirical study of Rab. In the procedure of the deduction, we adopt a statistical technique. This model takes account of not only the effect of congestion avoidance phase, but also the effect of slow start phase on TCP throughput, and also the effect of multiple random packet loss in one window, so this model is competent for actual Internet.