GridPP PMB Minutes 17 - 16th January 2002 ========================================= Present: Roger Barlow, Nick Brook, Tony Cass, Peter Clarke, Tony Doyle, Neil Geddes, John Gordon, Steve Lloyd, Robin Middleton. Full day meeting held at Imperial College, with members physically present. As this was a discussion meeting the Actions List was not considered; however it is appended at the end of these minutes for completeness. Tony Cass was welcomed as CERN liaison, replacing Les Robertson. John will update the email list and Steve will update the website. --------------------------------------------------------------------- Tony Doyle brought up the SCP4 problem and the Oversight Committee report. In response to the shortfall of 12M identified by SCP4, Dave will draft a 1 page document for SCP4 identifying where the 17M of GridPP money is being spent, and how it relieves the burden on the rest of the program through (1) the BaBar rebate (2) RAL staff and (3) RAL scalar computing hardware. A long consideration was given to the draft report from the oversight committee. Some errors were identified: Tony will collect and forward these. Points raised were: They ask for an Executive Summary in 12.c. We were unclear about what was meant. (Summary of the project, the overview, the explanation or what?) We are also unclear about what is requested as an 'Integrated Project' between CERN, GridPP and DataGrid, and are apprehensive about the feasability of anything of this nature. We are working on this as part of SC2 and the PEB at CERN. The 'dependencies' are dependencies between the experimental requirements and the items the DG releases contain. Unfortunately we do not know ahead of time what each release will contain. In practice experiments will use, or try out, whatever is provided in each release and give feedback. We will need feedback in March of how the experiments got on with the testbed; what bits they are using and what they think of them. This is how we decide when / if our High Level deliverables are being met. The EB should arrange. It is not clear what will happen to GridPP at the end of the project (assuming it succeeds). Considerable responsibility may be given to system managers (who may need protection in the coming Grants round.) We need to know the outcome of the next Spending Review before planning this area. To identify risks and the management of risks one has to look at the Gantt chart for the project. Tony will take this on. Our metrics for success are the experiment data challenges. Wisely or not, we had to allocate all our funds early. We note that we now have no contingency (except for PPARC) It was not clear what would happen to the money granted to groups for posts if a post became vacant and there was a gap before replacement: would the unspent funds revert to the group or the project? Neil will find out. In response to 12i: the TB and the new posts provide the glue between the experiments and the middleware. Our website needs improving. Steve and Andrew McNab should have another go at it. (12k) To coordinate JIF and SRIF sites (12l) Peter Clarke would encourage these sites to put up Grid software. We considered appointing a Project Manager, and decided after some discussion that the functions of such a manager are already adequately covered. Tony will meet the Astrogrid project manager to see what we can learn. Peter Clarke and the TB 'own' the architecture. ------------------------------------------------------------------------- Steve Lloyd reported on the consequences of extending the RAL supported posts from April to September 04. This is necessary to complete the project, though it removes our contingency. He would email the CB to get them to approve (it is not clear they have any alternative.) What these people do for the grid (Testbed as well as Tier 1/A) should be specified in the RAL SLA. (There is no need for a separate SLA). This needs clarification (especially with distributed Tier 2 sites as part of the grid) - John and Robin will consider. Computer Science posts funded by PPARC but through EPSRC should be in our accounts and reports. Relevant WP managers can tell them what to work on. ------------------------------------------------------------------------------ John Gordon will revisit the Tier 1/A staffing plans, linking in the eScience centre. Tenders for the spend this year look good, and the 900K should be spent by the end of the FY. Plans to buy 150 Dual P3 CPUs, 50TB of disk and 50 TB of tape were agreed. It was possible that the increased BaBar activity might saturate the 622MB connection from RAL to Janet. RB to investigate. The testbed will roll out incrementally, starting at sites with WP6 experts, then at other sites with support from WP6 experts. Queries will be channelled through the eScience centres. This structure was approved. It was asked whether GridPP would pay for machines dedicated to testbed work, and the answer was no. -------------------------------------------------------------------------- Few inputs had been volunteered for the Cambridge meeting. Suggestions should be made to Tony. The website is now being set up. Costs (room hire and incidental catering) are being split 60:40 between GridPP and the Cambridge eScience centre. WP reports whould be of highlights, not a complete list of everything. Technical talks are found more interesting than political ones. Roger and Nick will report on behalf of the experiments. --------------------------------------------------------------------------- Peter Clarke reported that the TB now exists officially. Steve will put the names on the website. Quarterly submissions to Datagrid had been reviewed. WP5 is getting back on track, as is the Liverpool half post. This shows that remediation mechanisms are in place and work. Pete has a spreadsheet that shows effort and deficiencies thereof. He will forward it to Dave, who needs it and has volunteered to process it. This should also include those working at RAL under the 'Grid' heading, as well as the institutes. -------------------------------------------------------------------------- Robin reported on the EU WP progress In WP3, MDS and FTREE seem to be converging. MDS and R-GMA are not and a decision will have to be taken between them. The TB will ask the WP manager (and Fab) for evaluation criteria to do this. For WP4 it is unclear what happens if LCFG has no long-term future WP5 recovery is back on track with 4 people in post. Is this enough to satisfy the oversight committee? Opinions should go to John, Robin and Peter. Andrew McNab has taken over as WP6 Liaison from Dave Newbold. The committee thanked Dave for all his excellent work. Concern was expressed that the TB and EB should be more closely coupled with WP8 to ensure better reporting. The biologists (EBI) should be invited to Cambridge to tell us about WP9 Recruitment is in progress though there are still some gaps. New posts will be put on the website by John. Note that EU datagrid pays travel for anyone working on DataGrid, not just those funded through them. This (plus some spend on desktop computing) is reimbursed on payment of deliverables. Fab wants to know how PPARC is controlling the 'unfunded' effort. Robin will produce a piece of paper. Intergrid want a progress report from BaBar by 17th February. RB will provide. ------------------------------------------------------------------------- Neil reported that CERN want a UK delegate on the Software and Computing Committee (SC2). John Gordon was the obvious choice. ------------------------------------------------------------------------- Tony (Cass) reported that recruitment at CERN was proceeding. The first RTAG was now set up. Hardware is being installed. there will be a workshop at CERN on March 11-15. In discussing relationshops between projects it emerges that the LHC computing Grid is the exploitation of the EU datagrid. -------------------------------------------------------------------------- Nick reported that the EB was entering a new phase now that the posts were in place. Its role is to ensure that GridPP is experimentally driven, and to act as a channel of communication among the experiments, and between the experiments and GridPP. The next EB will be mid March. ----------------------------------------------------------------------- The next PMB meeting of this type will be at Bristol on March 19th. Nick will make arrangements. ======================================================================= Actions: Outstanding and new ACTION 15.2: TD to iterate with Oversight cttee on understanding management of the Tier-2 centres, ACTION 15.3: SLL will write to the directors of each of the eScience centres on how best to work together. ACTION 16.2 NB and RB to plan Experiment session. ACTION 16.3 JG to plan Deployment session. Initial thoughts are presentations from all sites who have implemented EDG Testbed1 plus CERN, followed by discussion on future progress of EDG and GridPP testbeds. ACTION 16.4 PC and RM to plan Middleware session including feedback from Middleware Training Day on 31st Jan. Action 16.5 TD to contact Cambridge about local speaker(s). Fourth meeting is planned for May in Manchester but date not yet fixed. ACTION 16.6 RB ACTION 17.1 JG to update email list to replace Les with Tony ACTION 17.2 SL to update website to replace Les with Tony ACTION 17.3 DB 1 page document for SCP4 detailing where the 17M is going ACTION 17.4 ALL read Oversight Committee draft and report errors to Tony ACTION 17.5 TD Collate errors and forward to Guy Rickett ACTION 17.6 NB and RB Arrange for Experiments to report on testbed experiences in March. ACTION 17.7 TD Examine Gantt chart and identify risks, and consider how to manage them. ACTION 17.8 NG Find out who gets the money if there's a gap in a post. ACTION 17.9 SL with Andrew Mcnab; improve website. ACTION 17.10 PC Encourage JIF and SRIF sites to put up Grid software. ACTION 17.11 TD Contact Astrogrid project manager and see what we can learn. ACTION 17.12 SL Get CB to agree to extend RAL posts for final 6 months in 04 ACTION 17.13 JG and RM consider what 'running the grid' means. ACTION 17.14 JG revise Tier 1/A staffing plans. ACTION 17.15 RB investigate whether network from RAL to Janet may need upgrading and report to PPNCG. ACTION 17.16 ALL give Tony ideas for the Cambridge meeting. ACTION 17.17 SL Put TB names on the website. ACTION 17.18 PC forward spreadsheet to DB (including RAL grid effort and some words to explain meanings) ACTION 17.19 DB process spreadsheet from PC and use in reporting. ACTION 17.20 PC through TB: get evaluation criteria to make MDS/R-GMA decision. ACTION 17.21 ALL send feedback to JG, PC and RM about WP5 recovery. ACTION 17.22 RM pass on vote of thanks to Dave Newbold. ACTION 17.23 PC and NB improve communication of TB and EB with WP8 ACTION 17.24 RM invite EBI to give talk at Cambridge. ACTION 17.25 JG put RAL posts on website ACTION 17.26 ALL get grid certificates and learn how to update the GridPP website ACTION 17.27 RM write document describing how PPARC controls unfunded EU effort and send to Fab. ACTION 17.28 RB Provide Babar+Intergrid slides for Robin. ACTION 17.29 NG Tell CERN John Gordon is our delegate on SC2 ACTION 17.30 NB and RB Arrange EB meeting mid March. ACTION 17.31 NB arrange PMB meeting on 19th March in Bristol.