Breaking Boundaries - TACC as a Unified Cloud-Native Infrastructure for AI and HPC

Explore the innovative TACC (Tencent AI Computing Center) solution for managing AI infrastructure in this conference talk by Peter Pan from DaoCloud and Kaiqiang Xu from Hong Kong University of Science and Technology. Discover how TACC bridges the gap between Kubernetes and Slurm-based HPC setups, offering a unified cloud-native infrastructure for AI and high-performance computing. Learn about the five-year journey of implementing TACC at HKUST, supporting over 500 active researchers since 2020. Gain insights into the seamless user interface for job submissions, multi-tenant resource allocation strategies using CNCF HAMi and Kueue, and the robust distributed infrastructure featuring networked storage and RDMA via CNCF SpiderPool and Fluid. Understand how TACC addresses the challenges of managing large-scale GPU clusters for AI models while providing enhanced management granularity, stability, and usability for AI researchers.