Seminar - Persistence and Big Memory for Large-Scale Computing Systems
Supporting the below United Nations Sustainable Development Goals:支持以下聯合國可持續發展目標：支持以下联合国可持续发展目标：
Generative AI, LLMs, and many emerging large-scale computing systems continue to push memory demands into the terabyte-to-petabyte range. Big memory aggregates distributed memory into a unified address space. It has received significant attention from both academia and industry due to its salient features of high performance and large capacity. In this talk, we first revisit the evolution of system and architectural designs to elaborate on the fundamental principle underlying big memory, and then introduces state-of-the-art techniques to unlock its full potential. Specifically, we argue for a unified, byte-addressable Big memory fabric spanning local and disaggregated resources — delivering near-DRAM latency at massive scale via technologies such as CXL and RDMA. Our key claim is: persistence is not a bottleneck — it's a performance enabler: by shortening the critical path between compute and non-volatile storage, we eliminate the costly data shuffling that traditional hierarchies impose, bringing many benefits including throughput improvement, lower memory overhead, reduced operation latency, and reduced energy use.
Dr. Steve Liu is a Full Professor of Machine Learning, and a Full Professor of Computer Science at Mohamed bin Zayed University of Artificial Intelligence. (MBZUAI). He is also a Full Professor in the School of Computer Science, and a Full Professor (on a Courtesy Appointment) of Mathematics and Statistics at McGill University, where he has been a faculty member since 2007.
Dr. Liu has held a range of leadership and administrative roles across both academia and industry. He served as Vice President of R&D, Chief Scientist, and Co-Director of the Samsung AI Center in Montreal, where he led research and development of AI innovations across telecommunications, mobile computing, IoT, and embodied AI. He was also Chief Scientist at Tinder Inc., where he directed research and innovation for one of the world’s largest dating and social discovery platforms, valued at over $10 billion. In academia, Dr. Liu served as Associate Vice President of Research at MBZUAI, where he led research strategy, operation, and innovation. He also held the Samuel R. Thompson Chair Associate Professorship in the Department of Computer Science and Engineering at the University of Nebraska–Lincoln. Earlier in his career, he worked at Hewlett-Packard Labs in Palo Alto and the IBM T. J. Watson Research Center in New York.
Dr. Liu is an IEEE Fellow, and a Fellow of the Canadian Academy of Engineering. He is an associate member at the Quebec AI Institute (Mila), and McGill Center for Intelligent Machines (CIM). He was the chair of ACM SIGBED from 2021-2025. His research interests focus on AI/Machine Learning, Sustainable Computing, IoT, Financial AI, and Cyber-Physical Intelligent Systems. He has authored/co-authored 6 books/monographs and published over 400 research papers in major peer-reviewed international journals and conference proceedings, and received over 10 best paper awards from IEEE or ACM. He has served as an associate editor/advisor for several international academic journals and as a member of the technical or organizing committees of over 100 international conferences/workshops. Dr. Liu is also a serial entrepreneur and has advised/co-founded several high-tech startups.