From d.britton@IMPERIAL.AC.UK Sat Apr 14 17:36:05 2007 Date: Sat, 14 Apr 2007 12:23:05 -0400 From: David Britton Reply-To: GridPP Tier 1/A Board To: GRIDPP-T1B@JISCMAIL.AC.UK Subject: Input for Monday Dear Tier-1 Board members, As input to the Tier-1 Board meeting on Monday, here is a rough baseline for the Tier-1 hardware plans for this year. Plan-A: ~£1.1m for 1.5PB of disk to take the total disk to ~2.1 PB by April 2008. (The Disk requirement was 2.5PB and the PPRP approved 85%, or 2.1PB. I note that Babar's move to the Tier-2s gives us some surplus here, but that this is eaten up by phase-outs of older equipment during the year). ~£0.5m for 3.0MSI2k of CPU to take the total CPU to ~4.3 MSI2K by April 2008. (The CPU requirement was 5.1MSK2K and the PPRP approved 85%, or 4.3MSI2K. Again, there is some additional cushion freed up by the BaBar move but phase-outs of older CPU during the year decrease the total before the next purchase kicks in). ~£0.6m for Tape facilities giving 2.5PB of tape capacity, 700MB/s, by April 2008. (The Tape requirement was 3.2PB and the PPRP approved 85%, or 2.5PB) ~£0.1m for non-capacity and miscellaneous Tier-1 hardware. Total = £2.3m. Plan-A may not be possible because the figure I have from STFC for FY07 Tier-1 hardware is £1.9m. The main change here is that the new Tape Robot now appears in FY07, whereas in the GridPP3 submission it appeared in FY08. However, there are some funds available within GridPP2 which might not have been factored into this calculation, so this remains to be negotiated when the whole GridPP3 requirements for FY07 (and beyond) have been established. In terms of a Plan-B, we may be helped by delays in the LHC startup. One possibility is to split the disk purchase and perform the second tranche in April 2008 so that the cost falls in the next FY. It does seem unlikely to me that we will need all the disk on-line by April 2008; however, this has to be balanced against the risk of procurement problems; and the extra work involved in two procurements. Similar arguments might be applied to delaying the expansion of the Tape facility until FY08. In terms of progress, we need to converge on a best estimate of the requirements profile in the April 2008 to April 2009 period. On the one hand this represents the most expensive hardware we are going to purchase and we should not buy too early; On the other hand this is the largest fractional capacity-step and we should not underestimate the importance of delivering this on time. In parallel with this, the detailed GridPP3 planning needs to continue and then the profile negotiated with STFC once all the constraints are known. I would anticipate that this will be resolved by mid/late May and thus we will need to have made our best estimate of profile requirements by early May. Regards, Dave.