PMIx Tools WG - Meeting 9 - January 12th 2022

14 views
Skip to first unread message

Isaías Alberto Comprés Ureña

unread,
Jan 14, 2022, 5:39:32 AM1/14/22
to PMIx Standard Tools Working Group
Hello everyone,

I held up posting these minutes a bit, trying to find Dominik's master thesis in our public TUM website.  I will instead post an update here once it is available.

Important for those that were not able to join: 
We will move to monthly meeting, first Wednesday of the month, same time.

Participants:
Ralph Castain
Simon Pickartz
Thomas Moschny
Isaias Compres
Josh Hursey

Organisational:
- Generate a single link invite format for our Zoom calls
- Change to a monthly cadence
  - Next meeting: February 2nd 2022 
  - First Wednesday of the month after that
    - Reminders via the mailing list

Topics:
- Our WG is moving from brainstorming to active prototyping

- Slurm integration via Ralph's Slurm fork
  - Status: Ralph added hooks, notes are included
      - Kept up to date with Slurm upstream
      - Look for: rhc64 commits
      - Build system out of version control

  - Relevant thesis at TUM: Dominik (he participated at some PMIx in the past)
    - Dominik discussed the use of current PMIx apis vs. adding transactions at the PMIx level
      - Check the discussion on future work and conclusion near the end
      - A direct link to the thesis will be provided once published on TUM's website
    - Work based on process set updates
      - In the MPI sessions WG, this is one of the proposals being considered
      - There is a second proposal discussed in our previous call (Meeting 8)
        - This one is based on Session updates, instead of pset updates
      - There may be new proposals emerging, including Ralph's
        - Our focus is to be flexible and independent of programming model APIs, such as the MPI Sessions APIs

- Entering experimentation phase with transactions
  - Driven by DEEP-SEA research around malleability for scientific and engineering simulations
  - Transactions will be general purpose, not malleability specific

Reply all
Reply to author
Forward
0 new messages