CDF Status of Computing Donatella Lucchesi INFN and University of Padova Sep 6 2006 Donatella Lucchesi 1 Introduction Status and Plans for data and Monte Carlo production @FNAL Status and Plan Monte Carlo production @CNAF/EU Data outside FNAL CDF-Italy needs Remote shift control room (Fabrizio Scuri) Sep 6 2006 Donatella Lucchesi 2 Data Production Status Offline reconstruction algorithms frozen in fall 06 Data reconstruction performed for 1 fb-1 Post shutdown: Dataset cut-off: Sep 01 ~ 200 pb-1 Start producing new data Oct 01 Done by end of November Calibration process will be improved for the next step User level: Many analysis use st-ntuples produced by the production Farm (not standard yet for B physics) Diskpool in use for user (st-)ntuple, not production yet (st-)ntuples stored also at CNAF Sep 6 2006 Donatella Lucchesi 3 CDF Resources Needs as IFC Proportional Model: scale FY2005 resources FY200x demand = (FY2005 resources)x(FY200x/FY2005)data vol. Input parameters: Results: Sep 6 2006 Donatella Lucchesi 4 Where CDF can find new Resources Working on strategies to improve resources utilization: One-pass production Centralized ntuple production Run dependent Monte Carlo production Access GRID Resources Merge several GlideCAF into one:NAmCAF • Born as Fermilab-San Diego CMS group project • Now include: FermiGrid, San Diego, MIT, Florida lcgcaf: • CAF reimplementation to access LCG middleware • use gLite resource broker • CDF-Italy and INFNGrid project developed by: F. Delli Paoli, D. Jeans, S. Sarkar, I. Sfiligoi and D.L. Sep 6 2006 Donatella Lucchesi 5 NamCAF It start being used Sep 6 2006 Donatella Lucchesi 6 Resources available to LcgCAF Italy: INFN-GRID: 2MspecInt2K, extra CNAF ~700KspecInt2K Already included: Barcelona, Lyon, Glasgow, Karlsruhe will be presented Joint Physics meeting and very soon in production! Sep 6 2006 Donatella Lucchesi 7 Monte Carlo Production Plans cdf-UI job LcgCAF job Need to be developed for CDF in GRID output cdf-cnaf-1 data Robotic Tape Storage Sep 6 2006 EU GRID output data GRID Storage Element output data SAM station Donatella Lucchesi … storage validation Buffer area area 8 LcgCAF Characteristic & Performances CDF code distribution: AFS was a limitation, now Parrot access to all sites GRID failures due: sites misconfiguration, temporary authentication problems and services overloading in Resource Broker Use GRID retry and the CDF internal resubmission mechanism Samples Sep 6 2006 Events on tape with 1 Recovery B->Ds Ds-> 1969964 ~92% ~100% B->Ds Ds->K*K 2471751 ~98% ~100% B->Ds Ds->KsK 1774336 ~92% ~100% Donatella Lucchesi 9 Data analysis @CNAF An automatic (almost) procedure to store at CNAF: o B data in any form: BCHARM dataset and/or skimmed datasets and/or Bs-ntuple o st-ntuples (any format) for high-pt analysis Provide users running on data at CNAF with standard CAF (icaf) area, access via CAF tools and rootd Keep use GlideCAF until LcgCAF has same performances Use LcgCAF to access data: SAM-SRM interface development @FNAL. First deployment @CNAF: SRM installation (F. Delli Paoli) Test of native GPFS (D. Jeans) Lot of effort now in understanding which SRM Sep 6 2006 Donatella Lucchesi 10 CDF share respect to other experiments Total cpu time since April `05 Total cpu time August 2006 Need to agree with Tier 1 and Tier2 a minimum guarantee for CDF when LHC will start Sep 6 2006 Donatella Lucchesi 11 CDF needs: disk space CDF has some disk space assigned at CNAF. We ask for a little more to do data analysis Physics needs Use year Assigned B needs Asked Integrated (TB) (TB) (TB) 2005 32 35 tot CDF total T1 New CDF Integrated (TB) (TB) % 19 51 507 10% 2006 90 67 22 112 1056 11% 2007 130 96 31 161 1540 10% 2008 170 134 49 219 3960 6% Currently: 75 TB Sep 6 2006 Donatella Lucchesi 12 CDF needs: cpu’s Use year CDF Total T1 GR1 maggio 05 KSI2K 2005 494 2006 903 2007 1032 2008 1161 CDF % 1818 3420 5040 10800 27% 26% 20% 11% Dettaglio suddivisioni previste ALICE ATLAS CMS LHCb Totale LHC nudo Contingenza LHC Totale LHC con conting. 2005 CPU (KSI2K) Disk (TB) 220 30 320 45 350 110 110 50 1000 235 Tape 2006 CPU (KSI2K) Disk (TB) 330 132 480 192 525 210 165 66 1500 600 300 120 1800 720 Tape CDF is not asking more than what already discussed and approved. Tier 1 1000 seems235 happy! BABAR CDF AMS MAGIC ARGO ZEUS VIRGO Tot NON LHC Sep 6 2006 375 60 740 80 32 2 20 1 47 30 40 0 50 10 1304 183 Donatella Lucchesi 650 900 32 20 47 120 150 1919 187 90 3 1 30 20 331 13 Manpower Fermilab: I. Sfiligoi + San Diego group: “CAF +GRID transition” Replaced by D. Benjamin + R. Borgatti K. Genser T. Kuhr: “Data Handling and SAM” G. Compostella: Jul.-Sept. 2006, “CAF +NAmCAF” Italy: D. Lucchesi (coordination, SAM in Italy, data import-export) F. Delli Paoli (LcgCAF resp., CDF code, SRM development) ends in September 06 D. Jeans (GlideCaf-disk-user interface) new job Oct. 06 S. Sarkar (GlideCAF, LcgCAF, SRM) July 06 L. Brigliadori (support CNAF management) end January 07 New Assegno di ricerca @CNAF (Bando 10 settembre) F. Scuri +D. Fabiani Remote CO shift (Pisa) Interest express by colleagues for MC production coord. Sep 6 2006 Donatella Lucchesi 14 BACKUP Sep 6 2006 Donatella Lucchesi 15 Lcgcaf design Developers: F. Delli Paoli, D. Jeans, S. Sarkar and I. Sfiligoi advices Sep 6 2006 Donatella Lucchesi 16 Need for Physics Center Disk space: - BCHARM dataset and skimmed data are on disk - size of these datasets is scaled using luminosity, logging rate and event size Year 2005 2006 2007 2008 lumi (fb-1) Peak L3 rate N events BCHARM Skimmed total B (MB/s) x1e9 (TB) (TB) (TB) 1.40 35 2.0 23 12 35 2.20 60 3.4 54 13 67 3.80 60 5.7 77 19 96 6.10 60 9.2 107 27 134 What we need for B Data Sep 6 2006 Donatella Lucchesi 17