2014 06 05

• Who: Mark, Matteo, Shantenu, Ole, AndreM, Antons

• Agenda:

  • open TODOs
    • DONE AM: re-check 4096 ceiling
    • TODO AM: micro benchmarks for RP
    • DONE MS: MPI agent...
    • TODO AT: start on Cray agent, based on ATs scripts
    • TODO OW: get quantitative EnMDTK requirements
    • DONE AM: re-check with Matteo on exact problem
    • INV. AM: support MS on MPI agent?
    • TODO AM: address port forwarding ... (mom-nodes, head-nodes, ...)
    • TODO OW: check with Sergeij on Cray/BlueGene issue
    • DONE AT: put archer scripts into repo
    • TODO DINESH: check out port forwarding solution from Mark
  • MS-7 checkpoints:
    • May 8:
      • OW: OK simple MPI support for Stampede complete (prototype)
      • AT, OW: ?? draft architecture for Cray agent
    • May 15:
      • MS: ?? implementation proposal for MPI support beyond stampede
      • AM: ?? MPI integration tests set up
      • OW: ?? first prototype of non-MPI agent for cray
      • ALL: ?? agree on implementation plan for Cray agent
    • May 22:
      • AM: ?? MPI support integrated in Troy
    • May 29:
      • MS: ?? MPI support complete
      • OW/AT: ?? prototype for Cray agent -- reconfirm implementation plan
      • ALL: ?? documentation of performance bottleneck
    • June 5:
      • AM, OW: ?? integration tests for MPI stable for multiple use cases / machines
      • ALL: ?? agreement on optimization targets
  • status reports
  • ticket prioritization / milestones
  • (?) what role plays scheduling on agent level?

• Notes:

  • AM: tickets, tests, integration, 4096 checks

  • TODO AM: triage, this is probably not #104

  • TODO OW,DINESH: HT-BAC as scalability testing for MPI

  • Mark:

    • order of module load matters :( preexec...
    • TODO MS: document in resource configs
    • multicore and mpi support working
    • switched by resource config file
    • should be moved to CU level (#156)
    • removed multiple exec workers (no partitions)
    • multi threads for data remain...
    • merge with Cray track: ignore for now...
    • TODO MS: preexec does not work across nodes...

  • Ole:

    • resource configuration revamped
    • TODO OW: check

  • Antons:

    • Archer down...

  • TODO AM:

    • workdir

  • release:

    • TODO MS: sort tickets into tutorial milestones

• will data be merged?
• will 156 done?
• put data on agenda again