wiki:InfrastructureWGMeetingD20091123
Last modified 9 years ago Last modified on 11/23/2009 03:57:28 PM

We will be having our regular bi-weekly Infrastructure WG telecon on Monday, November 23, at 12 Noon CT.

Agenda

  • TG Allocation Update
    • Current Allocation
      • Allocated: 30K SUs on abe; 30K SUs on lincoln (nVidia Tesla GPUs)
      • Used: 0 SUs (abe); 0 SUs (lincoln)
      • Remaining: 30K SUs (abe); 30K SUs (lincoln)
      • How to get access: contact MikeF
    • Next TG Allocation Cycle
      • Proposals due Jan15 for Apr1 allocations
  • InfraWG Tickets
    • For the current list of open tickets
    • How to get access:
      1. Go to http://jira.ncsa.uiuc.edu and self-register
        • Very easy: Only your name and email address is needed
      2. Contact MikeF to get authorization to the LSST tickets
    • Note to lurkers: There exists a generic read-only account that can be used access these tickets. Contact MikeF for the password.
  • DC3b Storage Options/Costs?
    • Met with SET last week (discussed on DC3 call)
    • Meeting with PI on tomorrow (Tues)
    • Request in to TG allocations of PT1
    • Request in to NCSA allocations for PT2, PT3
    • Note to JeffK: Best guess is we will need some level of funding for infrastructure for PT2,3 (to be defined)
    • Unresolved from the DC3 call:
      • How are we serving the DC3b data?
  • NCSA Meeting tomorrow (Tues)
    • Initiation of regular ongoing collaborations with PI, Facilities, on status, current sizing estimates, design and planning, etc.
    • Initiation of a review of our current infrastructure design and costing by SMEs across NCSA.
    • Raise awareness of LSST, our HPC needs (DC3b), and drive issue resolution, validate assumptions, confirm policies (ACB mass stor, green PCF)
  • Cost Sheet Update
    • Baseline verion is v45
    • Current version now v67
    • Summary of Changes
      • LSST-52 fix misalignment of years between Requirements tab and Arch* and Base* tabs (-3992K/-4806K)
      • LSST-52 DR1/DR2 same year - don't buy DR1 HW 2 yrs early; also adjusted 2 yr ramp up (1/3 then 2/3)
      • LSST-49 New SiteProfile tab (in the Cost Sheet) and Key Technical Metrics section (in the Cost Summary sheet) (no cost change)
      • LSST-23 ArchAOS and BaseAOS: Slow Disk: Do not assume 1000GB per drive as a constant into the future (-190K/-664K)
      • LSST-24 Arch DPS and Base DPS: Fast Storage: Do not assume 500GB per drive as a constant into the future (-1147K/-3422K)
      • LSST-51 Floorspace tab: Floorspace calculation does not include floorspace for med storage for databases
      • LSST-56 P&C calculation does not include med storage for databases
      • LSST-59 The Power and Cooling and Power Cooling Lease tabs are redundant (and conflicting): Delete the Power Cooling Lease tab
    • Upcoming Changes
      • LSST-50 Floorspace tab: Floorspace calculation does not take into account the increase in drive capacities over time
      • LSST-10 Update Power & Cooling at Base Site (info already received from RonL)
      • LSST-36 Update Power & Cooling at ArchSite
    • Next steps with cost sheet
      • Full review each of the elements of the cost sheet (boxes of the mapping document)
        • More readable description of the formulas being used
        • Identification and documentation of assumptions
        • Identification and documentation of external data input
      • Serves two significant purposes
        • Allows for better internal reviews (validation of models and information used)
        • Provides justifications for external reviews
      • Results in an updated (or replacement of) Document-1684 and related documents ("Explanation of Cost Estimates")
  • Sync-ing up hardware model projections between Cost Sheet & diskIO (doc 1990)
    • drive capacity projections (double every 2 years, or every 3 years)?
  • Cost Estimate for ForcedSources
    • See LSST-9 for latest discussion
    • Question: I have seen a number of references to ForcedSources protrayed in a way that clearly implies that ForcedSources is already part of the baseline design. Should we treat it that way in the cost model?
  • Storage Model Costs for Maintaining CSS Database Performance
    • See my analysis document attached to the email announcing this meeting
    • i.e. the cost of going to min. # spindles for CSS database storage to meet performance requirements
    • Summary of current estimates
    • Large costs associated with this - need validation from group - error in sheet? revise model? change performance requirements? alternative technologies?
  • Current InfraWG Priorities
    • NCSA Review of Cost Sheet
    • DC3b hardware
    • Cost Sheet Tickets Above
    • Explanation of Costs Document
    • For the the complete list of all planned work, see the list of open tickets (link given above)

Notes

Attendees: Chris, Jacek, Daniel, Ray, Mike

  • TG Allocation
    • K-T [in absentia]: Used: 0 SUs is likely to persist for at least a few weeks until the code starts coming together. I think it's unlikely that we will have useful consumption numbers from DC3b itself before Jan. 15, so if those are needed to submit a proposal we'll have to come up with some early code we can run.
    • Mike: If we want more SUs on Apr 1, we need to build a case by Jan 15.
  • DC3b Storage Options
    • Mike: I have not heard back yet from my emails - I will follow up.
  • Serving DC3b Data
    • We need a statement of goals/requirements before the InfraWG can proceed in any meaningful kind of way in understanding the implications for infrastructure needs, options, and costs. Things to consider include separate servers, read-only access, response-time requirements, plus of host of other stuff.
    • Mike: I will raise during a DC3 call.