BaBar Grid. Tim Adye Particle Physics Department Rutherford Appleton Laboratory. PP Grid Team Coseners House 8 th November 2002

Transcription

1 BaBar Grid Tim Adye Particle Physics Department Rutherford Appleton Laboratory PP Grid Team Coseners House 8 th November th November 2002 Tim Adye 1

2 Talk Plan BaBar distributed computing model RAL Tier A Remote job submission BaBar VO and Authorisation Metadata Data distribution 8th November 2002 Tim Adye 2

3 The BaBar Collaboration 9 Countries 74 Institutions 566 Physicists 8th November 2002 Tim Adye 3

4 PEP-II e+e- Ring and BaBar Detector LER (e+, 3.1 GeV), I(e+ )=2.1 A Linear Accelerator HER (e-, 9.0 GeV), I(e- )=1.0 A BaBar PEP-II ring: C=2.2 km BABAR May 26, 1999: 1st events recorded by BaBar

5 BaBar s Distributed Computing Model Goal is to spread computing load much more around the collaboration Simulation production is already distributed 75% in the UK! Now have three new Tier A centres Lyon Objectivity (database) analysis (since last year) RAL Kanga (ROOT microdst) analysis (from May 2002) Padova Reprocessing (just starting) Also several Tier C sites (ie. Universities, 9 in UK) Analysis data format (Kanga vs Objectivity) is a matter of heated debate at the moment Whatever the future of Objectivity, Kanga (championed in UK/Germany) looks set to continue 8th November 2002 Tim Adye 5

6 RAL Tier A UK MoU with BaBar reduces our common fund contributions in exchange for providing Tier A facility RAL has now relieved SLAC of all Kanga analysis Impressive takeup from UK and non-uk users See Andrew s talk It is the primary repository of Kanga data ~20 TB on disk BaBar analysis environment tries to mimic SLAC so external users feel at home Grid job submission should greatly reduce this requirement 8th November 2002 Tim Adye 6

7 Remote Job Submission Short term (this month!) Allow SLAC or University users to submit BaBar analysis jobs to RAL or Lyon Tier A sites from their home machines dg-job-submit Simplifies local development and debugging, while providing access to full dataset and large CPU farms RAL vs IN2P3 selected explicitly by user canned JDLRequirements; dataset selection left to user Why couldn t we do this a year ago? BaBar authorisation (see later) Gatekeeper needed to be able to submit to production farm Define which BaBar configuration files to send with job Developed a procedure to merge all tcl files into one Resource Broker reliability better with EDG th November 2002 Tim Adye 7

8 Remote Job Submission Medium term (early next year) Allow remote submission to UK Farms and SLAC In principle this is already set up Select site (CE) based on user requirements Eg. Dataset available, software release, etc. Split job between sites based on available datasets Already have demonstrator for a canned analysis job 8th November 2002 Tim Adye 8

9 BaBar VO and Authorisation Use certificates from EDG and ESnet CAs for authentication Authorisation required to identify BaBar users Provides access to BaBar-specific facilities and environment Cannot maintain grid-mapfile by hand Doesn t scale to users Use existing SLAC BaBar user registration User provides certificate id at SLAC Automatic procedure checks AFS group and fills VO CEs use VO for authorisation Naturally handles people leaving the experiment 8th November 2002 Tim Adye 9

10 Analysis Metadata Currently have about a million Kanga files in a deep directory tree Need a catalogue to facilitate data distribution and allow analysis datasets to be defined. SQL database Locates ROOT files associated with each dataset Selections based on decay channel, run range, beam energy, reconstruction processing version, etc. Each site has its own (MySQL or Oracle) database Includes a copy of SLAC database with local information (eg. files on local disk, files to import, local tape backups) Some use of SRB for local Objectivity metadata at SLAC and Lyon 8th November 2002 Tim Adye 10

11 Data Distribution Kanga and Objectivity distribution currently handled by homegrown procedures Use bbftp. bbcp soon. Will look at GridFTP Next step is to run transfers using Grid job submission Web control pages under development Authorisation done using Grid certificates Looking at SRB and RLS for data distribution 8th November 2002 Tim Adye 11

12 Summary BaBar already has a highly distributed analysis environment RAL Tier A saves BaBar! Want to use Grid job submission tools now Looking at SRB and RLS 8th November 2002 Tim Adye 12