GridPP PMB Minutes 370 F2F (10.12.09) ==================================== Present: David Britton (Chair), Sarah Pearce (remote), Pete Clarke (remote), Robin Middleton, John Gordon, Jeremy Coles, Roger Jones, Steve Lloyd, Andrew Sansum, David Kelsey, Glenn Patrick, Dave Colling, Tony Cass, Neil Geddes, Tony Doyle (D Kelsey, Minutes) F2F @ Imperial College, London ------------------------------ 1. Welcome from DB ================== DB advised that the aim was to work today on GridPP4 proposal planning. We must have a good draft by the next face to face meeting in Glasgow (Jan 15th). 2. UK eScience review at RAL ============================ JG reported on yesterday's UK eScience review at RAL. Some feedback would be available in January. An invitation-only Town Meeting would be held on 9th Feb. 3. List of discussion documents =============================== Tier 1 Operations support (aka Deployment) Dissemination Experiment Support Middleware Support Project Management Tier 2 NGI Cloud Computing 4. NGI document =============== It was noted that Robin had an action to work on NGS plans - this had stalled because Andy Richards was out of the office. They will meet tomorrow to take further. GridPP3 currently funds people in security (2.3 FTE in operations, policy and vulnerability) and GOC operations (1.5FTE). There was discussion on the role of these in GridPP4. JG pointed out that EGEE-III funding for a number of people would end in May 09 (management, GOC, Tier 1, Tier 2 deputy coordinators, Dissemination). This was a problem for the last year of GridPP3 (need to come back to this another time). GridPP4 should not bid for NGI training and dissemination. There would also be no bid for service provision (CA and VOMS). These should be done by NGS. 5. Tier-2s ========== An approach had to be agreed today for presentation to the CB tomorrow. DB had been working on a document for the CB. The CB needed to agree the strategy and possibly also the process. DB presented some draft text. The strong steer received from the experiments was to support a core of well- found sites. CMS: The sum of Tier 2s was as important as the Tier 1 (especially for analysis). PMB agreed with this as a general statement. PMB believed that a minimum of 2 people was needed to run a successful Tier 2 site. ATLAS and CMS both need a minimum number of sites (size of machine room, power considerations etc). Current effort is 13 FTE Sysadmin, 4 Tier2 coordinators and 4 deputies (17 GridPP, 4 EGEE). Starting point was that existing effort (21 FTE) is the minimum viable. Staff at the core sites will need to take on shared support roles to cover the sites with smaller effort. Agreed that we will go for the 2 fte EGI-funded posts. It was agreed that the responsibilities need to be well defined. Effort allocation has to be done now as part of the bidding process, which is still a long time ahead of the start of GridPP4. First allocation of hardware funding was likely to be in the second year of GridPP4. PC stressed that physicists at non-core sites must not be dis-advantaged. PMB agreed to the overall approach. DB would send slides to CB at lunchtime. 6. Technical support ==================== A Cloud computing paper was presented, which had previously been dsicussed at an earlier PMB. Ł10k extra money for a pilot study as required. If successful we would then top-slice Ł100k. Middleware support was next addressed. Security was as previously discussed. Re Networking: GridSite was not part of the gLite consortium in EMI. TD needed to contact Andrew McNab to see what the situation was. Service Discovery is part of the EMI bid - this needed matching funding from NGS, OMII, but not GridPP4. Storage, Management and Replication: Currently GridPP3 funds 3 posts. It is at the heart of everything we do. We should work with the experiments to develop a strong case. Multi-cores: This could include modifications to the CE and needed to be closely tied to the experiments. This may be better tackled as part of the sys admin effort. The meeting broke for lunch. 7. Tier-1 ========= AS presented slides. Work in progress. Looking at any possible constraints. Machine room space, power and cooling are all OK, assuming constant funding. May need water cooling to racks. Electricity costs have been estimated. Disk cost extrapolation - slow down in growth in perpendicular recording and new technology unlikely before 2014/15. TK10C drives available by Q3 2010. Plan for which types of media are required needs the capacity plans. Non-capacity costs likely to increase from Ł150k to Ł300k per year. DB concerned that this is very large. Staff effort - plan to bid for all effort (GridPP + ESC). This is currently running at 24.6 FTE currently funded by GridPP, ESC and EGEE. This is not extra effort. It is currently all being used. Discussion ensued. Concerns expressed by the large number of posts involved. DB stressed that for all GridPP4 posts, we needed to prepare a case (half page) saying why it is needed, why this institute, and what would be lost if the post is cut. 8. Deployment support ===================== JC presented a paper showing a list of current Tier 2 co-ordinator duties and deputies. In future JC would like to allocate specific responsibilities to named individuals. This has to come out of the effort available to the Tier-2 sysadmin team. The name Tier-2 coordinator will not be used in GridPP4. JC's post comes under operations. SP proposed that all technical support and deployment support were linked together and called the "Operations Team". PMB agreed. This would include the team of core administrators (1 per core site), JC, security, network, data/storage. We should preserve the current Tier 2 structure. We still need some aspects of the current Tier-2 co-ordinator roles, (roaming support, quarterly reporting, etc). These roles can still be defined. 9. Experiment Support ===================== GP presented a paper. There was a need to continue the existing experiment support posts. GP proposed keeping these separate from the Tier-1 lists. We need good justification for the LHCb Tier1 post (Tier-1 very important, no other LHCb post funded, unlike ATLAS and CMS). Other recommendations: All LHC experiments should be consulted on optimising their Tier-1 storage- experiment links for GridPP4, i.e. data/storage links between Tier-1 and experiments needs to be continued. Should experiment support and operations be lumped together? No. We need separate work packages but referencing each other. 10. Knowledge Exchange ====================== SP reported that Neasan O'Neill's dissemination duties were well understood. It was considered important to continue with strong Outreach (as today). We need also to address Knowledge Exchange and Economic Impact. This is currently less well defined. One possibility for GridPP4 would be to drop dissemination, but this would be difficult. PMB agreed that we should bid at the current level. SP proposed increasing to 1 FTE to cover dissemination/outreach and KE. Agreed that we should bid for a full 1 FTE. GridPP4 financial model: There was no point in starting until we had an agreed list of posts. SP has the RAL staff cost estimates. No input yet on Tier-1 hardware, travel. SL raised the issus of FEC costs. All posts should be bid for including FEC. DB has input from the experiments on their requirements, but these are not yet official and are based on different assumptions. DB hoped to have figures for SP next week some time. 11. Next steps ============== It was understood that we had not yet reached the point where we could create a skeleton and assign writing tasks, but we did know we needed to justify posts. Experiment core sites needed to be proposed. Operations team tasks needed to be defined. DB went through the GridPP3 proposal and argued why much of that can be removed or reduced. The background documents do not need to be finalised - they were for discussion. Work could start on some sections of the proposal now. By the F2F in Glasgow we need to discuss a document on 15th Jan, i.e. need submissions by the start of that week (11th Jan) with a PMB that day. 11. AOB ======= There had as yet been no discussion regarding the structure and management of the project. Which boards did we need? DB suggested this has been well honed and there is no need now to change much. We needed to add the Operations Team. The Deployment Board, however, was not really required. We did however need meetings between the Operations Team and the PMB (twice a year). The User Board needed to continue. SP would produce and maintain a list of GridPP3-funded posts to track the transition into GridPP4. The financial model would be developed in time for the F2F in Glasgow. There would also need to be another CB meeting in January.