The 18th International Conference on Networking, Architecture, and Storage (NAS 2026) will be held from October 30 − November 1, 2026 at WuHan, China. The International Conference on Networking, Architecture, and Storage provides a high-quality international forum to bring together researchers and practitioners from academia and industry to discuss cutting-edge research on networking, high-performance computer architecture, and parallel and distributed data storage technologies. NAS 2026 will expose participants to the most recent developments in the interdisciplinary areas.
Each submission can be either a regular paper (up to 8 pages) or a short paper (up to 4 pages), including all text, figures, tables, footnotes, appendices, references, etc. Authors are invited to submit previously unpublished work for possible presentation at the conference. The program committee will nominate best papers for recognition in the three conference topic areas. All papers will be evaluated based on their novelty, fundamental insight, experimental evaluation, and potential for long-term impact; new-idea papers are encouraged. All accepted papers will be published in the Lecture Notes in Computer Science (IEEE) series and indexed by EI. Selected and extended papers will be recommended for journal publications.
- Abstract Submission Due: July 8, 2026
- Paper Submissions Due: July 15, 2026
- Notification to Authors: August 15, 2026
- Early-bird Registration: September 6, 2026
- Camera-ready Paper: September 4, 2026
- Conference: October 30 - November 1, 2026
- Submission Site: https://easychair.org/conferences/?conf=nas2026
Scaling 100K-GPU Interconnects: Status, Opportunities, and Challenges
Abstract
Next-generation super-AI is pushing clusters from thousands to tens of thousands—and even hundreds of thousands—of GPUs. In this exascale computing era, FLOPs are abundant, but bandwidth is not: moving data has become the critical bottleneck, rivaling the compute power of the GPUs themselves. In this talk, I will first examine the industry requirements, current landscape, and fundamental limitations of today's AI cluster interconnects. I will then present advances from academia, industry, and our own research in high-speed interconnect technologies. Finally, I will distill key lessons learned and outline directions for future research.
Biography
Professor Keqiu Li is Dean of the College of Intelligence and Computing and Dean of the Shenzhen Institute at Tianjin University. He is an IEEE Fellow and CCF Fellow, and also serves as Director of the Tianjin Key Laboratory of Advanced Networking and Director of the ACM China Tianjin Chapter.
His research focuses on emerging networking architectures, datacenter networks, software-defined networking, IoT, cloud computing, and mobile computing. Professor Li has published extensively in leading conferences and journals, including SIGCOMM, NSDI, INFOCOM, EuroSys, IEEE TC, TMC, and TPDS.
Professor Li has led major national and provincial research projects, including the National Key R&D Program of China and the NSFC Distinguished Young Scholars project, and has received several prestigious awards such as the First Prize of the Tianjin Science and Technology Progress Award, the First Prize of the Liaoning Technology Invention Award, and the First Prize of the MOE Natural Science Award.
Towards Edge AI Native Service Platforms: Rethinking Runtime, Deployment, and Migration
Abstract
Edge AI is shifting from isolated inference tasks to long-running services that coordinate model pipelines, data streams, service state, and accelerators near users and physical environments. Cloud-native and edge-native platforms are a useful starting point, but their main control objects–containers, nodes, links, and enrolled sites–are too coarse for such services. In this talk I present eAI+, a framework for edge AI native service platforms that make AI service graphs, edge resource fabrics, and participant contracts first-class inputs to platform control. eAI+ is organized around runtime, deployment, and migration, aiming to maintains AI service quality under latency, privacy, reliability, cost, and participation constraints. In addition, an incentive and plug-and-play participation module, named PolyLink, manages contributor onboarding, contracts, metering, reputation, rewards, and exit events.
Biography
Professor Jiannong Cao is currently the Otto Poon Charitable Foundation Professor in Data Science, Chair Professor of Distributed and Mobile Computing in the Department of Computing at The Hong Kong Polytechnic University (PolyU). He is also the Vice President (Education) and Director of the Institute for Higher Education Research and Development (IHERD) at PolyU. He served as Head of the Department of Computing, Dean of the Graduate School, and Head of the College of Undergraduate Researchers and Innovators (CURI), and the founding director of Research Institute for AIoT and University Research Facility in Big Data Analytics at PolyU.
Professor Cao is a member of Academia Europaea, a fellow of the Hong Kong Academy of Engineering, a fellow of IEEE, a fellow of the CCF, and a distinguished member of the ACM. He served as the Chair of the Technical Committee on Distributed Computing of the IEEE Computer Society from 2012 to 2014. In 2017, he received the Overseas Outstanding Contribution Award from the China Computer Federation.
