Improving Google’s BBR for Reduced Latency and Increased Fairness

BBR is a low latency congestion control algorithm developed by Google as an alternative to the traditional lossbased TCP (Transmission Control Protocol) algorithms, such as CUBIC. However, recent work shows that BBR suffers from 1) additional queuing delay due to competing flows and 2) poor fairness among flows with different Round Trip Times (RTTs). In this paper, we derive an analytic model for BBR’s Multiplicative Increase/Multiplicative Decrease (MIMD) sending rate control that reveals the origin of these issues. Inspired by Google’s recent patch for reducing queuing delay, we further improve BBR for lower latency. Moreover, our analysis suggests that applying a RTT-based adaptive pacing gain to probe bandwidth helps mitigating the fairness issue. Finally, we validate these improvements through simulations using the ns-3 network simulator.