Date

Mar 19, 2014 @ at 10:30a PDT (10:30a MST).

 

Tucson Location:  LSST Conference Room

Videocom: IP 140.252.24.8 (Direct Dial)

Call 1-866-330-1200, Participant Number 860-9352#

Attendees

  • Kem, Peter, George, Steve, Srini, Cathy, Lynne

  •  

Agenda

    1. MAF Test install update - Iain
    2. New machines - Srini
    3. Transition to git
    4. OpSim Dev:  Rolling Cadence implementation plan - Francisco
    5. OpSim Dev:  Speed Improvement plan & results from SLAC HPC - Srini
    6. JIRA process - ideal set up for scrum board
    7. JIRA planning - Abi/Cathy

Discussion Items

TimeItemWhoNotes
3min3Srinifit: Transition to git is still in progress
30min1All

MAF: Iain is in progress with a single user install, with trial output in home directories. Discussion continues on differentiating multi-user from single-user environments. Srini will document use cases and implementation differences. Iain will continue progress on testing pemission settitings, document them, and move on to installing MAF in a common area on opsimblitz2. Iain will see about acquiring a 2Tb drive for preserving data from older generation machines. We will need to identify changes in our workflow when MAF comes fully online so we make changes accordingly. Right now, SSTAR only has two configuration settings (design and stretch), MAF will have many configuation possibilities and a plan for storing/accessing these by the team and by the science community will be needed. If permissions settings can not be resolved satisfactorily, we will need a Plan B - which may include moving back to single user, lsst.

30min2AllNew machines: Document answers to questions about requirements and capabilities along with assessments of possible solutions (new machines, various configurations, SLAC).
  1. is it clear how much advantage is memory over solid state?
  2. what are the memory requriements now?
  3. are extremely high end machines needed?
  4. is there a need for more machines?
  5. telescope model - lookup table code analysis
  6. what are the bottlenecks - what is needed to improve? Not RAM - fine to have one run per core and run multiple runs -   more problems than cores? (OPSIM-149)
  7. do a gazillion runs - do on a gazillion cores simultaneaously (production ) but for developers, the more quickly you can do a run the faster you can get the answer and make progress.
  8. what would it take to parallelize the code? how long would it take?
  9. do we need machines for the workshop?
5min-CathyCadence Studies: Kem is working on rolling cadence runs but only integer year durations are possible
 4FranciscoRolling Cadence Code plan: Francisco was not present - presentation deferred to next week.
2min5SriniSLAC-HPC: Srini has gotten access to HPC machines and is working on installing opsim codebase.
1min6,7CathyJIRA: We'd like a better way to show our progress more accurately. Cathy will email Brian Selvy for recommendations on how to set up scrum boards or other solutions. We need to have time during our meeting to plan the next week's activity.

Action Items

    • Francisco - update profile picture in JIRA
    • Srini - check into how to declare versions for SSTAR v4.0 and OpSim v3.0 in JIRA and close/release appropriately (work with Francisco)
    • George - compose response to item 7 in Report from the Review Committee and send to Abi.

    • Abi - compose response to Report from the Review Committee

    • Cathy - email Brian about progress tracking in JIRA
    • Srini - differentiating multi-user from single-user environments; document use cases and implementation differences