Description
High-performance computing (HPC) is undergoing a structural transition as major laboratories and commercial providers adopt cloud-native architectures to power the next generation of scientific discovery. Leadership computing facilities, most notably the U.S. Department of Energy (DOE) laboratories, are planning and deploying systems aligned with the AI Factory paradigm, several of which are being procured under the Genesis Mission. These systems are engineered from the ground up to ingest, curate, and analyze large-scale scientific datasets while coupling equation-based simulation with modern AI methods.
This article outlines some opportunities and challenges for HPC application teams when adapting their simulation workflows to cloud-native environments. Many of these opportunities build directly on advancements in the HPC software ecosystem, including portable programming models, improved packaging approaches like Spack, the emergence of open software foundations, such as HPSF, and the growing role of curated software distributions such as E4S.
