BBRv1 outperforms BBRv2 in CDN77's global QUIC delivery benchmarks
CDN77 conducted live experiments to optimize dynamic content delivery using the QUIC protocol, comparing the performance of Cubic, BBRv1, and BBRv2 congestion control algorithms across global Points of Presence. The study revealed that BBRv1 achieved a larger congestion window size (CWND) and outperformed BBRv2 in Europe, North America, and the Middle East, while BBRv2 performed better in Asia by maintaining a more stable connection window.
Key Takeaways
- Over 80% of clients establish new QUIC connections rather than utilizing stream multiplexing features.
- BBRv1 achieved superior congestion window sizes (CWND) compared to BBRv2 in three out of four global regions tested.
- BBRv2 showed better stability for dynamic content in Asia compared to both Cubic and BBRv1.
- Congestion control algorithm overhead remained negligible, with no measurable CPU utilization differences between Cubic, BBRv1, and BBRv2.
- The RFC-compliant Google Quiche library was used to benchmark data collected from Q4 2024 through Q1 2025.
Why It Matters
Optimizing congestion control within the QUIC protocol is critical for low-latency dynamic delivery where even minor packet loss triggers buffering. These findings suggest that the 'latest and greatest' algorithms like BBRv2 are not yet universally superior, often trading raw throughput for fairness in a way that may disadvantage specific regional streaming workloads. For CDN strategists, this highlights a shift toward region-specific network tuning rather than global default configurations. As the industry moves toward Media over QUIC (MoQ), the ability to maintain stable congestion windows will be the primary differentiator for live sports and interactive streaming platforms. Watch for BBRv3 adoption rates as developers attempt to bridge the performance gap between BBRv1's speed and BBRv2's fairness.
Additional Context
The transition to QUIC and HTTP/3 is accelerating rapidly across the B2B streaming landscape. Per Cloudflare data from October 2025, HTTP/3 adoption has reached 35% of global traffic, transitioning from an experimental phase to a production reality. This protocol shift is primarily driven by the need for better performance on lossy wireless and mobile networks. Research from November 2024 indicates that while traditional TCP-based HTTP/2 suffers from head-of-line blocking during packet loss, QUIC's UDP-based architecture allows independent stream processing, resulting in response time improvements of up to 47% in high-latency environments. The debate between BBR versions remains a core focus for infrastructure engineers. Per a 2024 report from the Applied Networking Research Workshop, BBRv2 was specifically designed to be 'friendlier' to legacy Reno and Cubic flows, but this fairness often comes at the cost of lower goodput in wide-area networks (WAN). Google has since introduced BBRv3 to address these throughput-fairness trade-offs. Simultaneously, the IETF's Media over QUIC (MoQ) working group is developing new standards to unify ingest and distribution. Companies like Akamai and YouTube joined the OpenMoQ Software Consortium in early 2025 to fast-track relay software capable of caching and distributing real-time media without traditional fragment delays. Furthermore, the commercial impact of QUIC is substantial. Per The Business Research Company in February 2026, the global QUIC-enabled CDN edge market is projected to reach $3.79 billion by 2026, growing at a CAGR of 33.1%. This growth is heavily concentrated in North America, currently the largest market, though the Asia-Pacific region is expanding the fastest due to massive mobile-first streaming populations. These market dynamics make the technical selection of congestion control algorithms like those tested by CDN77 a financial imperative for providers competing on quality of service.
Read full article at cdn77.com