Analytical Model of TCP Relentless Congestion Control

TL;DR

This paper presents an analytical model of the Relentless Congestion Control (RCC), a novel TCP congestion control variant, and its enhancement RCC+ with loss detection, to evaluate its network performance impact.

Contribution

It introduces the first analytical model of RCC and its improved version RCC+ with loss detection, facilitating better understanding and assessment of this congestion control scheme.

Findings

Model accurately describes RCC and RCC+ behavior

Simulation implementation in ns-2 validates the model

RCC+ reduces RTO caused by retransmission loss

Abstract

We introduce a model of the Relentless Congestion Control proposed by Matt Mathis. Relentless Congestion Control (RCC) is a modification of the AIMD (Additive Increase Multiplicative Decrease) congestion control which consists in decreasing the TCP congestion window by the number of lost segments instead of halving it. Despite some on-going discussions at the ICCRG IRTF-group, this congestion control has, to the best of our knowledge, never been modeled. In this paper, we provide an analytical model of this novel congestion control and propose an implementation of RCC for the commonly-used network simulator ns-2. We also improve RCC with the addition of a loss retransmission detection scheme (based on SACK+) to prevent RTO caused by a loss of a retransmission and called this new version RCC+. The proposed models describe both the original RCC algorithm and RCC+ improvement and would allow to better assess the impact of this new congestion control scheme over the network traffic.