ECP Supercontainers

The ECP Supercontainers Project was a flagship effort within the U.S. Department of Energy Exascale Computing Project (ECP), a multi-billion dollar national initiative to deliver exascale computing for DOE science and national security missions.

Andrew served as Principal Investigator from 2019 to 2023, leading a multi-institutional team across DOE national laboratories to develop portable, reproducible HPC software deployment through container technologies. Co-PIs included R. Shane Canon (LBNL), Reid Priedhorsky (LANL), and Sameer Shende (University of Oregon).

Key contributions:

• Investigated Podman as a community-supported, rootless container runtime for DOE leadership computing facilities, demonstrating scaling across thousands of nodes on NERSC’s GPU-accelerated Perlmutter system [1]
• Advanced Charliecloud R&D to minimize privilege required for building and running HPC containers, enabling fully unprivileged container builds compatible with multi-user HPC security models [2]
• Developed best practices, tools, and documentation for building and running containers on DOE leadership computing facilities (ALCF, OLCF, NERSC)
• Led container runtime integration with GPU-accelerated systems and facility workload managers (Slurm, PBS Pro, Flux)
• Organized tutorials and training workshops at SC and ISC, reaching hundreds of HPC practitioners annually
• Established the Supercontainers GitHub organization for open-source HPC container tooling
• Hosted and organized the DOE/E6 Container Working Group (2018–2021)

Related publications: [3], [4], [5], [2], [1]