The Bottleneck-Bandwidth and Round-trip (BBR) congestion control algorithm was introduced by Google in 2016. Unlike prior congestion-control algorithms (CCAs), BBR does not rely on signals that are weakly correlated with congestion (e.g., packet loss and transient queue delay). Instead, it characterizes a path using two parameters, bottleneck bandwidth and round-trip propagation time, and is designed to converge with a high probability to Kleinrock’s optimal operating point [34]. Essentially, in stable state, BBR maximizes throughput while minimizing delay and loss. Google has used BBR for a significant fraction of its network traffic both within its datacenters and on its WAN since 2017 [15]. BBR’s interaction dynamics with Cubic, the widely used CCA in the Internet, has received intense scrutiny: Some studies observed BBR to be unfair to Cubic, or generally loss-based CCAs. Google, to its credit, has diligently revised BBR’s design to address the criticisms. This paper focuses on characterizing the promises and potential of the third, and most recent, revision of BBR—introduced to the public in July 2023. We empirically evaluate BBRv3’s performance across a range of network scenarios, e.g., considering different buffer sizes, round-trip times, packet losses, and flow-size distributions. We show that despite the improvements and optimizations introduced in BBRv3, it struggles to achieve an equitable sharing of bandwidth when competing with Cubic, the widely used CCA in the Internet, in a wide range of network conditions.
Click here to read full paper.