January 7, 2010

Performance Meeting - 2010/01/07
Phone Number: (812) 856-3600
PIN:          002690#
Agenda:
  • Review architecture writeups
  • PEP schedule
  • Detailed schedule
Notes:
Attendees:  Shava (UCSD), Robert (IU), Thomas (TU-D), Warren (TACC), Mauricio (UF)
  • Writeups
    • Benchmark section - Robert will add a little text describing reasoning for the benchmark selection.
    • Instrumentation section - Shava made modifications - Netlogger2OTF should be possible but Thomas will need samples - Shava will forward netlogger sample to Thomas
      • Warren pointed out there may be other interfaces than Pegasus - e.g., Moab directly - also we may want to capture more detail from Moab that Pegasus will not know about
      • Shava will talk to Joe/Jennet to get more information on Moab logging
  • Performance tools
    • Group can provide packages and config recommendations - install done by who? (ask on sw call)
    • Should group provide images or can they be built on the fly? (ask on sw call)
    • Need to test configuration and installation order of packages to make sure they play well together - Thomas can help - we need a machine to test on - wait for new idataplex machine for IU
  • Usage information
    • Mauricio brought up on last call -- Shava will ask on sw call
    • Warren pointed out we may want to know usage of performance tools
  • PEP schedule - Shava reviewed and Vampir and PAPI activities scheduled to start out in Aug -- should be ok -- Inca milestones will need to be shifted since architecture still be defined
  • All committees asked to have detailed project schedule by end of month - Added first draft below -- Shava will make some revisions based on discussion today and ask folks to look at next week.  Look at time estimates and disregard start/end date for now.
  • Experiment harness interfaces -- no changes since holidays -- should mirror portal interfaces -- will revisit soonish