Reproduce “To Adopt or Not to Adopt L4S-Compatible Congestion Control? Understanding Performance in a Partial L4S Deployment” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Fatih Berkay Sarpkaya, Fraida Fund, Shivendra Panwar
This study from NYU Tandon School of Engineering rigorously evaluates the performance of L4S-compatible congestion control algorithms, like TCP Prague and an L4S-compatible BBRv2, in various partial L4S deployment scenarios. The research experimentally demonstrates that L4S flows frequently suffer significant performance degradation or unfairly dominate classic TCP flows when coexisting across diverse network queues, especially in single-queue ECN environments and with BBR variants.
Key insights
- Unilateral adoption of L4S-compatible congestion control by senders carries substantial risks, as L4S flows are highly likely to encounter settings where they are either harmed or cause harm.
- L4S flows often exhibit an 'either-or' outcome in partial deployments: either severely harmed (receiving minimal bandwidth) or unfairly dominating classic flows.
- DualPI2, the L4S-specific dual-queue AQM, is not well-tuned for coexistence with BBR flows, often leading to BBR dominating L4S flows in test scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.