meeting Mar 3 2022

Notes for 2022-03-03 meeting

• date & time: Thu 3 Mar 2022 - 14:00 CET (13:00 UTC)
  • (every first Thursday of the month)
• venue: (online, see mail for meeting link, or ask in Slack)
• agenda:
  • Quick introduction by new people
  • EESSI-related meetings in last month
  • Progress update per EESSI layer (+ GitHub App, MPItrampoline)
  • Monitoring EESSI infrastructure
  • 2021.12 version of pilot repository
  • AWS/Azure sponsorship update
  • Update on EESSI paper
  • S4 project proposal for NeIC call
  • Upcoming events
  • Q&A

Slides

• available at https://github.com/EESSI/meetings/blob/main/meetings/EESSI_meeting_20220303.pdf

Meeting notes

(by Kenneth)

Quick introduction by new people

• Elisabeth Ortega (HPCNow!)

EESSI-related meetings in last month

• EESSI community meeting in Amsterdam 15-16 Sept?

  • co-located with CernVM workshop (12-14 Sept)

• plan to set up proof-of-concept support for GPUs in EESSI pilot 2021.12

  • add script that should be run to EESSI repo (TODO Alan)
  • provide basic instructions on what to do
  • add cuda-samples installation to facilitate testing (WIP Kenneth)

Progress update per EESSI layer

Filesystem layer

(see slides)

• WRF input data (cfr. https://github.com/EESSI/software-layer/pull/170) is relatively small (<1GB)
  • temporary solution could be to add it to EESSI pilot repo
  • benefit is limited for just WRF
  • proper setup would require setting up a separate repo/external web server, etc. (see https://cvmfs.readthedocs.io/en/stable/cpt-large-scale.html)
  • use S3 bucket as storage backend?
  • this is mostly about easy access, not fast access
  • learn from CERN repositories

Compatibility layer

(see slides)

• ...

Software layer

(see slides)

• Terje: plans to extend PyGHee functionality to easily post comments back into PR that triggered event?

• experiments with MPItrampoline

  • some software (Score-P, mpi4py, CP2K) needed patching, mostly because they don't fully follow the MPI standard
    • most of those patches can be pushed upstream
    • CP2K patch is more complex, may be good to get feedback there from CP2K developers
  • MPItrampoline has only been tested on top of OpenMPI for now
    • enabling runtime switching to different MPI backends (MPICH, MVAPICH2) should be as simple as providing an MPIwrapper for these
    • being able to leverage a vendor MPI is more difficult (w.r.t. linker being used, location of files, etc.)

Monitoring EESSI infrastructure

• http://status.eessi-infra.org (not HTTPS yet)

2021.12 version of pilot repository

(see slides)

AWS/Azure sponsorship update

(see slides)

• separate EESSI data repository could be set up in Azure?
• CitC developer (Matt Williams) was given access to EESSI Azure credits for testing support for Azure in CitC that's being developed and contributed by Hugo (MS Azure)
• Should we also set up an EESSI login node in Azure?

Update on EESSI paper

• Already published o/: https://onlinelibrary.wiley.com/doi/full/10.1002/spe.3075

S4 project proposal for NeIC call

• 2nd attempt to get funding for EESSI through NeIC

Upcoming events

• project proposal for BioHackathon Europe 2022 (https://biohackathon-europe.org)
  • in the scope of ELIXIR project
  • suggested by Sebastian Moretti via EasyBuild Slack

Q&A

• ...