Archived edition. SNAI 2024 took place on 18–20 October 2024 in Hangzhou, China. Go to the current edition, SNAI 2026 →

Keynote Speakers

Two keynote talks delivered at SNAI 2024, Hangzhou, 18–20 October 2024

SNAI 2024 featured two keynote talks delivered by leading researchers in networking and learning systems. Both keynotes were held in plenary and streamed virtually to remote attendees.

KN

Why Networks Resist Learning

Kenji Nakamura · University of Tokyo · 18 October 2024

Despite a decade of effort, learned components still struggle to displace hand-designed congestion control, traffic engineering, and routing in production networks. This talk examined the gap between ML benchmarks and production network deployment: non-stationary traffic, multi-objective trade-offs, the absence of accurate simulators, and the cost of rare failures. Nakamura argued that progress will come from closer collaboration between networking and ML researchers, and from evaluation methodologies that share the operational constraints of operators. The talk drew on case studies from learned congestion control in data centres and from learned traffic engineering in WANs.

SW

Congestion Control After the Learning Turn

Sarah Whitfield · University of Melbourne · 19 October 2024

After a decade in which learned congestion control moved from research curiosity to deployed reality, the field has begun to converge on a small set of design patterns. Whitfield surveyed the state of learned congestion control and discussed what the next generation of protocols must do differently: explicit handling of fairness across flows, robustness to RTT heterogeneity, and integration with end-to-end transports designed for AI workloads such as distributed training. She closed with a position on the role of congestion control in networks whose primary traffic is no longer web or video but collective communication between accelerators.

Speaker biographies

Kenji Nakamura is a Professor in the Department of Information and Communication Engineering at the University of Tokyo, where he leads the Network Architecture Lab. His research interests span programmable networks, learned traffic engineering, and the operational deployment of ML components in carrier networks. He is a member of the SNAI steering committee and served as Program Chair for the inaugural SNAI 2024 symposium.

Sarah Whitfield is a Professor of Computer Science at the University of Melbourne. Her research focuses on transport protocols, congestion control, and the interaction between learned components and the underlying network. She has served on the program committees of SIGCOMM, NSDI, and IMC.