CB Meeting 14 - 21/01/10 ======================== DRAFT Minutes - Not yet approved Present: Peter Watkins (Birmingham), Nick Brook (Bristol), Peter Hobson (Brunel), Ian Bird (CERN), Andy Parker (Cambridge), Nigel Glover (Durham), Phil Clark (Edinburgh), Tony Doyle (Glasgow), Roger Jones (Lancaster & ATLAS), David Hutchcroft & Tim Greenshaw (Liverpool), Andrew McNab (Manchester), Brian Foster (Oxford), Steve Lloyd (QMUL & Chairman), Pedro Teixeira-Dias (Royal Holloway), Norman McCubbin (RAL), Neil Spooner (Sheffield), Antonella De Santo (Sussex), Chris Allton (Swansea), Jon Butterworth (UCL), Dave Colling (CMS), Glenn Patrick (LHCb), Dave Britton (Project Leader), Sarah Pearce (Project Manager). Apologies: Trish Mullins (STFC) 1. Approval of minutes from meeting 13 ====================================== The minutes were approved. 2. GridPP4 Proposal Introduction - DB ===================================== Slides introducing V6 of the proposal, circulated in advance of the meeting, were presented by DB. DB noted that clear guidance was obtained from STFC w.r.t. 20% reductions - tough decisions had been required and this was contained within the current draft proposal. DB also noted, however, that the current working allowance is higher than the planning figure. 3. GridPP4 Proposal Discussion ============================== Note comments are not necessarily in chronological order but have been group together by topic. AP suggests that on Impact there is an existing Impact Fellow at Cambridge who is potentially available at the 0.5FTE level. DB notes the useful suggestion and will take this offline with SP. AM comments that current version looks credible already at this stage. DB notes that high-level comments are welcome with the "eye of a reviewer". PH asks the approach taken w.r.t. fEC. DB responds that the current assumption is that all posts are research active. This be revisited if we are asked to de-scope. AM brings up Manchester's Tier-2 allocation in the context of RB's e-mail and DB and RJ responses. Why are the numbers not allocated to maximise numbers of large Tier-2s? What about existing very large sites? DB noted the importance of the prior strategy in defending all Tier-2 posts. e.g. IC and Glasgow have been identified as centres of excellence. More generally the distinction is to identify "well-found Tier-2 analysis" sites. Beyond that, the aim is to engage as fully as possible with many Tier-2 sites although the defence for this approach is relatively weak. There is no correct answer. TG asks whether this takes account of existing hardware? SL responds that the issue is not CPU, nor even just disk, but disk throughput. RJ notes existing hardware reflect simulation capacity, not analysis analysis. SL adds that the criteria now address future capability. AM notes that M'cr bought new hardware using DPM to address their distributed dCache deployment problems. RJ notes even after this change M'cr still failed to meet all three of the ATLAS analysis criteria (M'cr currently meets one of the three criteria and hence are identified as an amber site). AP notes that it is hard to manage sharp changes in manpower numbers. DB responds that overall funding is sufficiently dire that a strong defence of each post is needed. TG suggests moving the impact factor post to Manchester but DB notes that this will not address the Tier-2 problem. DB notes that M'cr's previous (GridPP3) large mapower allocation to Tier-2 was justified based on potential future delivery. NM notes that the change is in 15 months hence there is time to manage the change. NM is however concerned that a rather large resource is not being used as effectively as possible. DB notes that such arguments also apply to other potentially large sites e.g. Bristol. JB notes that UCL is in this category but is willing to accept the proposed outcome. UCL still plan to deliver for GridPP4. In response to NB, DB notes the plan to reward performance during GridPP4 with effort. The proposed response is a mid-term review and the ability to redistribute posts at that point. DB notes, however, that the same issue of managing change between sites will arise. PH feels that a risk is necessary but also that the post allocation must be fixed at this stage (esp. to manage 1/2 FTE funding). DB asks the CB to approve the 3+1 year funding model for Tier-2 posts. AP notes that 1/2 FTEs mean that hard numbers are needed to manage effort locally i.e. a 3-year minimum is required. DB notes that all grants would be issued explicitly noting the mid-term performance review. AM notes that the time period and performance criteria will be needed in advance. SL notes that the criteria are all related to "performance with the real data". DB notes that the overriding aim is to deliver what the experiments want. DB adds that GridPP will work in advance of the evaluation period to establish appropriate metrics. TG notes that all grants work by identifying good performance and these posts are no different. PC notes that "value for money" is also important TG adds that being seen to be fair is also crucial, with agreed criteria. DC endorses SL approach of "meeting expt. needs". PS notes that the mid-term review process will however be difficult. DB notes that the three previously identified (expt-VFM-site) criteria do differ, but they will be approached in this order. External reviewers are anyway likely to say there are still too many sites, hence further rationalisation may be needed. PC notes that a 4-year position advertised as 2(+2 subject to performance) can be attractive locally to manage delivery. PC adds that retaining that tension w.r.t. individual performance is important. SG asks if the planed period of assessment to assess global performance extends over 3 years. SL responds that we will need to weight towards the end of a given period. MC production metrics are fine, we first need to establish analysis metrics. SG responds that these metrics are needed early in the process (in order for all sites to respond to them). SL notes that the target is to have these agreed in 2011. SL asks the CB to endorse the proposed approach taking the current Tier-2 post distribution, with a mid-term review, leading to 3+1 year post allocations. General consent: the CB endorses this approach. SL asks if the CB is happy that we go forward with the current overall draft as is? General consent: the CB endorses this approach. SL and DB note that the plan is to circulate next Thursday's version for OC also to the CB. DB re-iterates that identifying high-level weaknesses in arguments is important in the current version. PC notes that greater clarity is needed w.r.t. Tier-2 hardware. SL responds that the plan is to issue hardware in 2+2 year tranches. AP notes that a sensible procedure is needed for allocation of smaller hardware sums to sites. DB notes difficulty of managing this downstream. PH asks how confident we are of the existing GridPP3 hardware fund allocation since there is a need to tender now. DB responds that this is 90% probable. SP adds that she will be e-mailing all GridPP3 PIs shortly w.r.t. the most recent Tier-2 hardware. NM notes that student effort on CMS is mentioned in the document and being handled offline.