Self-adaptive Scheme to Adjust Redundancy for Network Coding with TCP

Abstract

Network coding has emerged as an important potential approach to improve the robustness and efficiency of data transmission over lossy wireless network. TPC/NC protocol proposed by Sundararajan et al incorporating network coding into TCP by online coding, TCP/NC has the advantage of naturally adding network coding to current network systems and masking non-congestion packet losses from the congestion control algorithm, However, in TCP/NC the values of redundancy factor R can’t be adapted based on the characteristics of the underlying channel.

In this paper we propose a novel self-adaptive scheme to dynamically adjust R based on the collective feedback information of ACKs, which contain the information of sinks decoding matrix. Since the scale of decoding matrix is indicators of the lossy channel condition, the source adjusts R based on the channel conditions, avoiding unnecessary TCP rate reduction and preventing the network from entering in a congestion state. The TCP/NC with our self-adaptive scheme is realized in OMNET++. Simulation results over realistic network scenarios show that our scheme in conjunction with the standard TCP/NC significantly outperforms the previous redundancy approach in reducing size of decoding matrix , and produces better TCP-throughputs than the standard TCP/NC, TCP-Reno.