BBR-R: Improving BBR’s RTT Fairness by Dynamically Adjusting Delay Detection Intervals

Proposed by Google in 2016, the Bottleneck Bandwidth and Round-trip propagation time (BBR) congestion control mechanism has been widely used on the Internet. BBR reduces packet loss and significantly minimizes end-to-end packet latency, garnering substantial attention in recent years. However, recent studies have shown that BBR suffers from severe fairness problems when flows with different Round Trip Times (RTTs) share a bottleneck link. Existing methods for fairness improvement mostly limit the packet sending rate of longer RTT flows, which is not a good way since the root of the problem is not considered and their results still have room for improvement. In this paper, we propose a simple but effective modification to BBR, named BBR Refined (BBR-R), to improve throughput fairness among flows with different RTTs. Based on the root causes of unfairness, BBR-R adaptively changes the delay detection intervals for draining the queues in the bottleneck link buffer while maintaining link utilization. We evaluated the proposed method with some existing works including BBRv2 and confirmed that while ensuring high link utilization and lower latency, BBR-R improved the fairness index by up to 40%.