2014 03 06

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

• Agenda:

  • Anton's perf test plan
  • expectations on data staging performance (compared to BJS)

• Notes:

  • perf test plan
    • AM: 1, 2, 3: very relevant
    • SJ: may need more CUs to reach steady state on large pilots
    • AM: notifications vs. state polling for large numbers of CUs? relevant for troy...
    • SJ: weak/strong scaling?
      • possibly address more cleanly later on
  • data staging expectations
    • OW: as good as manual threads + python + sftp
    • MT: compare to standard scp command
    • SJ: define set of scenarios
    • TODO AM:
      • define test plan
      • provide saga-python perf numbers
      • possibly delegate to Mark