Liverpool T Bowcock,G Patel,J Phillips,G Hughes,C Parkes
RAL G Patrick,C Brew,D Salmon
Glasgow/RAL P Soler
Oxford I McArthur,F Harris
Bristol N Brook
IC D Colling
CERN J Harvey,E van Herwijnen
A Khan(Edinburgh),U Marconi(Bologna),
A Tsaregorodtsev(Marseille),A Sansum(RAL)
Dave Salmon reported that Globus 1.1.3 was installed front-ending the CSF farm, and it had been used for the successful submission of jobs. It was expected that a 'standard installation specification' might come out of the Datagrid project, and this is awaited. Some experience was being acquired with the use of the FTP tools in Globus.(see slides from G Patrick below)
Eric van Herwijnen stated that 1.1.3 had been installed at CERN since mid July. However he had only managed to use it successfully in the past week, consequent to being allocated the appropriate access priveleges. We all awaited a statement from CERN/IT as to lanned support for Globus front-ending CERN farms. John Harvey commented that a public service was not being offered yet, but IT are commissioning ~5 machines to be used for GRID tests.
Girish Patel has successfully installed Globus 1.1.3 on one of the master machines front-ending MAP (hep52.ph.liv.ac.uk). ((http://lhcb-comp.web.cern.ch/lhcb-comp/grid/gpatel.ppt) Tests using globusrun and globus-rcp commands, submitted from accounts at RAL and Liverpool, had been successful.
(http://lhcb-comp.web.cern.ch/lhcb-comp/grid/gpatrick.ppt)
Glenn Patrick overviewed progress in the UKGRID. The community hoped to know by the end
of October the allocation of UK GRID funds to the Particle Physics community. A series of
technical meetings has commenced, exchanging experiences with GRID software(notably
progress at RAL,Manchester and QMW).
The organisation of a distributed testbed is being discussed (RAL, Liverpool, Glas/Edin,etc...). Also a hierachical management structure is being defined, going down from a PPARC steering committee to the technical sub-groups. CLRC organisation is about to be split into PP and e-science for all of CLRC.
(http://lhcb-comp.web.cern.ch/lhcb-comp/grid/frank-wp8-2.ppt)
Frank Harris reviewed the organisation and status of the Datagrid bid to the EU. The contract was to be discussed in Brussels on Sep 11, with a view to signing on Oct 25 for a 3 year contract, worth 10M Euro. This would commence Jan 1,2001 and the EU funds would be used for staff.
LHCb participate in the WP8 (HEP applications) workpackage, and had participated to WP8 meetings in Marseille to refine the EU proposal, and discussed how the 4 LHC experiments will work together. WP8 will be represented on an Architecturegroup (1 person/work package), and each experiment will be represented by 1 person on a Datagrid Technical Board. We will need to map our LHCb work programme to the somewhat generic milestones and deliverables of WP8, and overall WP8 needs to define, via the Architecture Group and Technical Board, how we relate to the Testbed and Netorking groups, as well as to the middleware groups.
(http://lhcb-comp.web.cern.ch/lhcb-comp/grid/tbowcock.ppt)
Themis Bowcock overviewed MAP successes, issues and possible improvements. MAP has been used for MC generation for lHCb, H1, Delphi, Atlas and CDF. This has been largely successful. Issues are the ease of use for the physicist user, the operating system version, storage, up-time and the use of Compass nodes. Other more significant issues relate to the hardware configuration (Disks,Power Supplies,Hubs vs. switches).
There are planned improvements covering all of the above areas, these starting in September, and continuing till December. Liverpool wish to merge in Globus/GRID activities with the production programme in a compatible, incremental way.
There are several challenges regarding the importing/exporting of data, and also the possibility of doing analysis 'sur place'. It was agreed that this analysis model possibility was something we want to test in the GRID programme. More generally we need to explore how Liverpool, as a Tier-2 centre, will fit into the the UK GRID, Datagrid and LHCb infrastructures.
(http://lhcb-comp.web.cern.ch/lhcb-comp/grid/jharvey.ppt)
The management of appropriate SICB versions for production will be managed by Eric(CERN), Martin(Liverpool) and David (RAL).
It had been defined by the LHCb physics group tha the high priority productions are Bbar inclusive in DST2 format without trigger cuts ( a sample of 500K to be done at RAL), and a sample of 2M Bbar inclusive (after L0/L1 cuts) hopefully to be produced at MAP. This latter sample will involve generating some 20M events since some 10% pass the cuts. This also involved merging some 2M minimum bias events. We need to check whether the 20M need to go through full or partial reconstruction. Assuming the latter(a minimal reconstruction) gives an estimate of some 6 months production on MAP. Themis stated that in principle Liverpool agreed , but it could not be 6 months full-time since MAP had other commitments. Liverpool would get back to LHCb a.s.a.p as to the fraction of MAP available from late October onwards.
It was agreed to use networking for the data transfers (estimated 60 KByte/s
requirement). Chris Parkes said he had obtained 500 kB/s running FTP jobs overnight
Liverpool-CERN. C Brew has made measurements of 400-500 kB/s RAL-CERN. It was agreed to do
tests Liverpool-RAL. For the production it was proposed to generate data onto MAP disks,
copy to RAL, and copy from RAL to CERN where the central bookkeeping
will be updated. This system will be set up by Eric, Martin and Dave.
A manual quality checking exists with histos viewable from WWW. This will be done at CERN where the production QC responsibility will lie (Liverpool with install the software but not use on a production basis).
C Parkes will inform Eric of the random number seed management scheme used by DELPHI.
The bookkeeping system will require a joint development between Eric and Liverpool. It is suggested that a new bookkeeping file be generated on the MAP master.
F Harris said all information regarding production and GRID activities will be put on LHCb GRID pages (including all meeting minutes).
It was agreed by all that we must organise ourselves to merge GRID software into our production system incrementally, starting in a simple way which will facilitate the uniformity of our user interfaces.
Eric and Frank will generate a short technical summary paper stating how we intend to organise our production at MAP.
It was emphasised that had been a general LHCb GRID meeting, and not a UK meeting (our French and Italian reps had sent their apologies). In this spirit we should aim to have our next meeting perhaps either at Lyon or Bologna.
Themis was thanked by all for his hospitality and organisation.