115. Challenges in HPC/AI infrastructure development: Cyfronet AGH perspective

Abstract

During the talk, I will describe the current state of HPC and AI infrastructures from the perspective of the latest Top500 and Green500 lists and I will present some challenges for the future development of large computing systems. Roadmaps of leading CPU and GPU vendors will be discussed, focusing on power and cooling requirements and some new issues that future architectures bring to HPC systems. The talk will also cover the development of EuroHPC Joint Undertaking supercomputers in Europe and present Helios – Poland’s fastest supercomputer is now being commissioned at Cyfronet.

About the author

Marek Magryś has been working at the Academic Computer Center Cyfronet AGH University in Krakow since 2006. He currently serves as Deputy Director for High Performance Computing. He has extensive experience in designing, implementing, testing and operating computing and data storage systems for HPC and AI applications, as well as optimizing computing applications for efficient use of storage resources. He was the chief architect of the fastest Polish supercomputers of recent years: Prometheus, Athena and Helios as well as coordinator of the infrastructure of the EuroHPC PL project which is building the next generation of computing machines, including those using quantum computers. He is also the manager of the EuroHPC National Competence Center in Poland.