An end-to-end flow control approach based on round trip time

Abstract

Traditional TCP flow control schemes, such as Tahoe and Reno, continuously increase the window size until possible packet loss is detected. These schemes generate congestion unnecessarily by themselves and utilize the buffer space inefficiently. Therefore, Vegas was proposed to accommodate the drawbacks. It uses the measured round-trip time (RTT) to obtain the information required for the flow control. In this work, we propose a flow control scheme, RTT-based TCP, which uses the RTT directly to control the window size adjustment. RTTbased TCP comprises four mechanisms: the modified exponential window increment, the RTT measurement, the queue occupancy estimation, and the quick queue-occupancy reduction. Simulation results show that RTT-based TCP outperforms Vegas in the aspects of throughput, fairness and efficiency of buffer usage.