University of California, Los Angeles (UCLA), United States of America
We present a practical approach for managing the scientific packages in the HPC cluster environment of a research university that has a diverse user base. The primary goal is to minimize the HPC operational team’s burden of installing and maintaining a large number of software packages to support the broadband spectrum of the university’s computational research. We propose a hybridizing management method that can harness the power of modern software management frameworks and the flexibility of in-house, or manual, installation sources, and at the same time present a coherent view to the end users. The Spack framework was used in this work. Our hybrid approach is applicable to using other framework tools. The manipulation of Spack-generated environment module files and the typical workflow are illustrated and discussed based on our use case.