Digital Twin for Datacenter: HPC4AI UniTO Case Study

The HPC4AI (High-Performance Computing for Artificial Intelligence) datacenter at the University of Turin’s Computer Science Department was established to meet the rapidly growing computational demands of interdisciplinary AI research. HPC4AI innovates by redefining the traditional roles of Cloud and High-Performance Computing (HPC) systems, where the Cloud provides a modern interface for HPC, and HPC acts as an accelerator for Cloud applications. To date, it has supported over 40 research projects spanning diverse fields such as astronomy, medicine, and human sciences. Additionally, HPC4AI serves as a research and development platform for exploring, developing, and testing novel datacenter technologies. It features a variety of experimental computing platforms and the first prototype of a two-phase evaporative server cooling system. This work outlines the operational management of HPC4AI, highlighting challenges, lessons learned, and key opportunities related to digital twins for datacenters.