[lustre-discuss] any benefit(s) in using I_MPI_FILESYSTEM_FORCE=lustre with Intel oneAPI MPI?

[Rob Kudyba]

Has anyone tested and/or recommend Intel's oneAPI MPI
<https://www.intel.com/content/www/us/en/docs/mpi-library/developer-reference-linux/2021-12/other-environment-variables.html#GUID-852BA712-230A-491F-8659-13ACC42AEB4A>
 module that purports to optimize MPI jobs by using these options:
I_MPI_FILESYSTEM=on
I_MPI_FILESYSTEM_FORCE=lustre

We'd like to know if this would improve performance for MPI jobs.
NASA references this as well on their Lustre best practice page
<https://www.nas.nasa.gov/hecc/support/kb/lustre-best-practices_226.html>.

Are there any suggested MPI jobs (links to any examples are appreciated)
that would make for a good before and after test?

I'd also be curious to see if Intel collaborated with anyone at Whamcloud
on this.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20240425/0c8af15b/attachment-0001.htm>