Providing Accessible Software Environments Across Science Gateways and HPC

Abstract

While High-Performance Computing (HPC) resources are powerful for tackling complex, computationally intensive analysis and modeling problems, access to these resources varies across disciplines. Domain scientists in a variety of fields such as social and environmental sciences often lack in-depth technical skills (e.g., familiarity with terminal, knowledge of job schedulers) to effectively utilize HPC resources, hindering desired research. In this context, CyberGIS-Compute is a middleware toolkit designed to democratize HPC access with the main goal of enabling domain scientists in diverse fields to solve computationally intensive problems. A key challenge facing model developers on CyberGIS-Compute is to create a containerized software environment for their models. Domain experts unfamiliar with HPC are generally unfamiliar with containerization technologies (e.g., Docker, Singularity) and thus unable to create/test containers to execute their models. But if they have access to science gateways, they would want to use these familiar software environments on HPC resources. This paper describes a novel approach to integrating the Cern Virtual Machine File System (CVMFS) into CyberGIS-Compute to provide consistent software environments across science gateways and HPC resources.