Development of a high-speed transport protocol with TCP-Reno friendliness

The traditional TCP-Reno is not capable of achieving a high throughput in high-speed networks due to its highly conservative and loss-dependent congestion control mechanism. Although TCP-Vegas achieves a good link utilization due to its proactive type of congestion control mechanism that uses RTT as an indication of network congestion, it lacks friendliness to TCP-Reno. Protocols such as HighSpeed-TCP and Scalable-TCP have been proposed to address this problem, but they behave non-friendly to TCP-Reno especially when sharing high bandwidth links. In order to improve the friendliness with TCP-Reno and throughput performance in high-speed networks, we propose a new congestion control mechanism, TCP-xHS (High-speed) that harmonizes the fast ramping and proactive congestion controlling approaches of HighSpeed-TCP and TCP-Vegas. Through simulation results we justify the effectiveness of TCP-xHS for a wider range of link bandwidths.