filmov
tv
ARCHER2 Webinar: The State of Containers
Показать описание
Michael Bareford, EPCC
It is anticipated that there will be an uptake in the use of containers for running research codes. It is hoped that containers will facilitate reproducibility and portability.
You may see a paper that presents some intriguing results produced by a numerical simulation. The paper references a definition file, allowing you to build the container (within which the research code runs) and deploy to a HPC platform of your choice. You find that you can reproduce the results. You then tweak the input settings to explore a number of what-if scenarios and in effect stand on the shoulders of the author(s).
Is this expectation realistic? I will present my attempts to run various research codes within containers on two HPC platforms, Cirrus and CSD3. Where possible, I investigate whether or not there is a performance penalty incurred by running containerised software.
It is anticipated that there will be an uptake in the use of containers for running research codes. It is hoped that containers will facilitate reproducibility and portability.
You may see a paper that presents some intriguing results produced by a numerical simulation. The paper references a definition file, allowing you to build the container (within which the research code runs) and deploy to a HPC platform of your choice. You find that you can reproduce the results. You then tweak the input settings to explore a number of what-if scenarios and in effect stand on the shoulders of the author(s).
Is this expectation realistic? I will present my attempts to run various research codes within containers on two HPC platforms, Cirrus and CSD3. Where possible, I investigate whether or not there is a performance penalty incurred by running containerised software.
0:54:38
0:50:06
0:53:30
0:55:03
0:48:50
1:04:55
1:02:59
0:18:43
0:43:45
0:49:21
0:57:33
0:56:49
0:56:37
1:01:34
0:36:10
0:43:52
0:27:54
0:51:57
0:27:00
0:39:07
0:50:32
0:42:28
0:49:22
4:00:38