Abstract
Containers are seeing widespread use in the world of High Performance Computing, with many HPC Centers either providing their own containerization solution or adopting existing ones like Singularity and Apptainer. The demand for containerization options come from users who want to take advantage of the portability and reproducibility containers can provide, as well as being able to build and use applications that are only distributed in container form or are otherwise unsuited to natively run in an HPC environment. The users served by the Oak Ridge Leadership Computing Facility are no exception. We go over the past and current containerization offerings at the Oak Ridge Leadership Computing Facility, mainly focusing on the Summit supercomputer. We arrive at using a combination of Podman and Singularity to allow users to build and run containers directly on Summit, without requiring external resources or hardware for any step of the process. We look at a couple of projects running on Summit that greatly benefited from being able to use containers on Summit. And we compare benchmarks running natively and in containers on Summit at different scales, observing minimal performance difference and consistent behavior across all tests.
Original language | English |
---|---|
Title of host publication | PEARC 2023 - Computing for the common good |
Subtitle of host publication | Practice and Experience in Advanced Research Computing |
Publisher | Association for Computing Machinery, Inc |
Pages | 273-277 |
Number of pages | 5 |
ISBN (Electronic) | 9781450399852 |
DOIs | |
State | Published - Jul 23 2023 |
Event | 2023 Practice and Experience in Advanced Research Computing, PEARC 2023 - Portland, United States Duration: Jul 23 2023 → Jul 27 2023 |
Publication series
Name | PEARC 2023 - Computing for the common good: Practice and Experience in Advanced Research Computing |
---|
Conference
Conference | 2023 Practice and Experience in Advanced Research Computing, PEARC 2023 |
---|---|
Country/Territory | United States |
City | Portland |
Period | 07/23/23 → 07/27/23 |
Funding
This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725.
Keywords
- MPI
- containers
- high performance computing