Session Date/Time: 21 Jul 2025 12:30

CCWG

Summary

The CCWG session at IETF 123 in Madrid covered updates on the hackathon, drafts regarding congestion window increase and BBR, and performance analysis of BBRv3. Discussions focused on practical implementation challenges, testing methodologies, and future directions for congestion control algorithms. Key themes included the need for more experimental and simulation data, especially in Wi-Fi and high-speed data center environments, as well as the development of better testing tools and methodologies.

Key Discussion Points

Hackathon Update: Progress on testing FQ Codal, FQPy, and RFC 8511 implementation, with a new AQM evaluation suite being launched in NS3. Community feedback solicited on incorporating the suite into the main NS3 distribution.
Congestion Window Increase Draft: Discussion on limiting congestion window increases when senders are rate-limited. Focus on refining the language around pacing mechanisms and their impact on the proposed rules.
BBR Update: Focus on simplifying the BBR draft and making it less TCP-centric. Open issues include defining BBR's behavior when coming out of idle, application-limited pacing text, probe RTT interval, and offload budget for QUIC. Specific asks for experimental and simulation data were made.
BBRv3 Performance Analysis: Presentation of experimental results on BBRv3's self-induced queuing delay, fairness issues, and behavior with real-world traffic. Discussion on the motivation for allowing higher queuing delays in BBRv3 and the impact of Wi-Fi jitter.
Implementation and Testing: Emphasis on the need for better testing methodologies, particularly for BBR and QUIC implementations. Suggestions included defining test scenarios, sharing implementation guidance, and leveraging tools like Qlog.
Probe RTT Interval: Discussed whether the Probe RTT interval should be 5 or 10 seconds. Agreement on the need for adapting the probe RTT to account for Starlink reconfigurations and dynamically changing paths.

Decisions and Action Items

Congestion Window Increase Draft:
- Remove the sentence about pacing "usually" limiting the number of packets per RTT.
- Authors to rephrase pacing paragraph based on feedback received.
- Working group last call to commence after editorial cleanup.
BBR:
- Authors to add an experimental section in the document covering the issues surrounding Probe RTT
- Authors to file an issue for adaptive pro RTT
- Working group to focus on producing guidance for implementation testing.
- Working group to collect and consolidate common "trip-over" locations from those implementing BBR.

Next Steps

Congestion Window Increase Draft: Authors will incorporate feedback, perform editorial revisions, and initiate a working group last call.
BBR: Authors will address open issues, focusing on the idle behavior, pacing text, probe RTT frequency, and QUICK offload budget. Experimentation and data gathering, particularly regarding the 5/10-second probe RTT interval, will continue. Community feedback and collaboration on testing and implementation challenges are strongly encouraged.
BBRv3 Performance Analysis: Documentation of the design background and trade-offs related to BBRv3's queuing behavior. Additional research on dynamics on real Wi-Fi links and high-speed data center links.
TSVWG: Stuart Cheshire to present L4S numbers in the TSPWG session on Wednesday.