BBR
BBR (Bottleneck Bandwidth and Round-trip propagation time) is a TCP congestion control algorithm proposed by Google. Instead of treating packet loss as a congestion signal, it actively measures a link's bottleneck bandwidth and minimum round-trip time, pacing packets accordingly — significantly improving throughput and latency on high-latency, lossy paths.
How it works
Classic algorithms (Reno, Cubic) assume loss means congestion and cut their sending rate exponentially on every drop, which cripples performance on paths with random loss such as cross-border links and wireless networks. BBR instead models two key metrics — bottleneck bandwidth and minimum RTT — and paces traffic toward their optimal operating point. Linux has shipped BBR since kernel 4.9, typically paired with the fq qdisc: enable it with sysctl net.ipv4.tcp_congestion_control=bbr and verify with ss -i or tc. For cross-border VPS links, video calls, and lossy mobile networks, enabling BBR is often the highest-value zero-cost network optimization available.
Examples
- 1.Websites and downloads on a cross-border VPS get noticeably faster, especially during evening peak loss
- 2.SSH sessions and video calls hold steadier latency without repeated stalls
- 3.Large transfers over lossy mobile networks finally saturate the available bandwidth