Congestion Control of HighSpeed TCP Using Learning Automaton

Abstract

In the Internet, TCP is a wide commonly used transport layer protocol. However, the utilization of bandwidth of conventional TCP decreases in the high bandwidth. As one of mitigation measures, HighSpeed TCP has been proposed. However, the utilization of bandwidth of HighSpeed TCP decreases when the bandwidth of bottle-neck-link changes, because the recomended value of parameters of HighSpeed TCP are determined assuming packet size is 1,500 Bytes and line speed is 10 Gbps. In this letter, we propose a new congestion control method introducing learning automaton into HighSpeed TCP, aiming to adjust the important parameters of HighSpeed TCP adaptively.