FEBBR: A Fairness-Enhanced Approach for BBR Congestion Control
Huijun Yuan, Yi Han, Yi Zhong, Mingxi Liao, Zhenhui Yuan, Gabriel‐Miro Muntean · 2023
The Bottleneck Bandwidth and Round-trip propagation time (BBR) is a novel algorithm used for TCP congestion control, designed to optimize both throughput and latency for web servers and datacenter backbone connections. Unlike traditional loss-based congestion control algorithms, BBR measures and considers bottleneck bandwidth and round-trip time (RTT). However, recent research has shown that TCP BBR suffers from severe fairness problems when coexisting with ordinary TCP flows. In this paper, we propose a Fairness-Enhanced BBR (FEBBR) to address the fairness problem. FEBBR handles the aggressiveness of BBR by introducing delay deviations. Extensive emulation experiments on Mininet demonstrate that FEBBR improves the fairness index by 20%-40% compared to BBR, while not affecting much the throughput performance. In addition, FEBBR reduces RTT by 27%-40% while improving RTT fairness by 39%.