Overcoming Fairness and Latency Challenges in BBR With an Adaptive Delay Detection

In 2016, Google introduced a new congestion control algorithm called Bottleneck Bandwidth and Round-trip propagation time (BBR). BBR aims to set the data-sending rate of flows to operate at an optimal operating point—maximizing throughput, minimizing delay, and reducing packet loss. However, when multiple BBR flows share a bottleneck link with a deep buffer, they often deviate from the optimal point, resulting in significant queuing delays, and serious fairness issues especially when flows with different RTTs are present. Existing methods to enhance BBR’s performance typically cap the sending rates of longer RTT flows to improve fairness and reduce latency, but these do not tackle the fundamental problem. In this paper, we introduce a modification to BBR, termed BBR Refined (BBR-R), which enhances throughput fairness among diverse RTT flows and lowers queuing delays. By targeting the underlying causes of unfairness, BBR-R adaptively adjusts delay detection intervals to clear queues in the bottleneck link buffer while preserving link utilization. Our experimental results show that BBR-R enhances RTT fairness by 30% over BBR and other variants, and reduces latency by 30% compared to the original BBR.