Linux Containers continue to gain momentum within data centers all over the world. They are able to benefit legacy infrastructures by leveraging the lower overhead compared to traditional, hypervisor-based virtualization.
After last year’s panel discussion within the ‘Invited Program’ this proposed session about containers within the This BoF will recap the challenges for containerized workloads in HPC (kernel-bypass, shared POSIX file-system, distributed computing) and provide an overview of the HPC container landscape by segmenting it in palatable slices.
1. How to run a containerized process in HPC?
2. How to build a container so that it is portable and performant?
3. How to distribute a container image across sites and projects?
4. What HPC specifics need to be taken into account?
Contributors:
- Shane Canon (Lawrence Berkeley National Labs)
- Carlos Eduardo Arango Gutierrez (Red Hat)
- Christian Kniep (QNIB Solutions)
- Andrew Younge (Sandia National Laboratories)
- Jonathan Sparks (HPE/Cray)
