1 / 11

Tier-1 Planning Version-35

Tier-1 Planning Version-35. David Britton Imperial College. Inputs: Money. GridPP1 Hardware – £2,314K

Download Presentation

Tier-1 Planning Version-35

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Tier-1 Planning Version-35 David BrittonImperial College Collaboration Meeting

  2. Inputs: Money GridPP1 Hardware – £2,314K Final out-turn now thrashed out with CCLRC. Less than previously reported because some money spent from the Hardware budget on other items such as staff costs, travel, overheads andgeneral GridPP operations costs. GridPP2 Hardware – £2,999K This is increased slightly from previously reported due to slow hires at CCLRC. This number will continue to fluctuate at the 10% level because it effectively includes the contingency - E.g. Tier-2 depreciation (there is currently 135K further savings from CCLRC posts but not added to HW budget). Profile The current model contains a profile which includes the current capacity purchase and a first stab at planning for Tape to be described later. There is slightly more bias towards disk since it is hard to over-allocate. Tier-1 Board

  3. Input: Hardware Costs From Sep. Meeting: Email from Andrew Sansum 2/Aug/05: Mid 2005 - 1.4 K/TB (recent quotation for small system with 400GB drives - 12 port PCI/RAID) 2007 - 0.63 (assume in early 2007 800GB drive - 1TB sometime in 2007 - 24 port PCI/RAID) 2008 - 0.48 (linear interpolate 2007-20010) 2009 - 0.32 (linear interpolate 2007-2010) 2010 - 0.168 (informed Guess 1Tb/inch - 3TB drives) Tier-1 Board

  4. Inputs: New Hardware Costs Disk costs are 33% more than expected CPU costs are 50% more than expected Tier-1 Board

  5. Outputs: Tier-1 Capacity Tier1Plan29 Tier1Plan35 ~25% reduction in Disk and CPU Tier-1 Board

  6. Output: TD’s LCG Pledge Table + Errors +/- 10% +/- 20% +/- 30% +/- 40% +/- 50% 2431 3587 4471 1233 829 1375 1815 2234 629 413 2043 3242 555 4364 313 Tier-1 Board

  7. Nominal “Fair-shares” as a fraction of Request For the LHC experiments and BaBar, the orange boxes are where we would fail to meet (80%) of International MOU commitments. Tier-1 Board

  8. Disk/CPU/Tape Tier-1 Board

  9. BaBar Fair enough, but buying an extra 100TB of disk now means ~150-200 TB less disk can be bought for 2008: - CMS/ATLAS already failing to meet MOU commitments from 2008. - CMS Tier-1 will be downgraded to a Tier-2. - UK physicists will be disadvantaged. - UK’s international reputation will be damaged. - Unknown common fund implications. - LHC will have real data. Argue for increase in resources from “fair-share” level: - to honour International MOU commitments. - to prevent Tier-A from being downgraded to a Tier-2. - to prevent UK physicists being disadvantaged. - to preserve/recover UK’s international reputation. - to enable common fund rebate (but $460k is hard to understand). - Babar has real data. Tier-1 Board

  10. BaBar My opinion: The only argument is a pragmatic one: BaBar has a problem right now which could be solved. Bringing spend forward compounds a problem that already exists in 2008 so it would be better to squeeze the LHC this year and manage within the currently planned capacity. Note that we have already mortgaged the “real-data takes precedence” argument because there is no significant allocation for the large Linear Collider or Neutrino communities in the present planning so in 2008 there is already no possibility of applying this argument. If funding is bought forward, then I believe the minutes should record the principle on which the decision was made. We are going to be in the identical situation next year (BaBar request is >300TB) and the year after with the LHC experiments. Tier-1 Board

  11. Issues • The latest purchase prices for Disk and CPU both cause a considerable shock to the planning and jeopardise the LCG MOUs. • Uncertainty in demand for Tape and risk of delays commissioning Castor2 suggest that we move cautiously to the new Robot. Would like an affordable plan that maintains 9940 infrastructure until milestones can be met with the TK10/Castor2 installation. • The BaBar issue is a close analogy to the situation foreseen with the LHC experiments, particularly CMS, in 2008. The issues are coupled. • The next purchase. Tier-1 Board

More Related