1. Amber Boehnlein, FNAL D0 Computing Model and Plans Amber Boehnlein D0 Financial Committee November 18, 2002

2. Amber Boehnlein, FNAL Computing Status DO has a highly successful computing structure in place  Sequential Access by MetaData (SAM) catalogs and manages data access  Robotic storage with reliable drives and media  Domino provides high I/O capacity and user access to large amounts of data  Commissioning the commodity backend  FNAL Reconstruction production farm processing increasing to 35 Hz  Basic software infrastructure in place  Fruitful collaboration with the Computing Division on joint projects.  MC generation performed at collaborating institutions  DORECO has basic functionality  Basic Filtering at L3  Online output rate is at design. The Building Blocks are in place… … we are producing first physics results

3. Amber Boehnlein, FNAL Computing Architecture Central Data Storage dØmino … … Central Analysis System Central Farm Remote Farm Remote Analysis Linux Desktops (ClueDØ) DØ

4. Amber Boehnlein, FNAL Data Flow Data Handling System Remote CentersCentral AnalysisClueD0Regional FarmCentral Farm Robotic Storage Raw Data RECO Data RECO MC User Data

5. Amber Boehnlein, FNAL Analysis Model dØmino … CAS Analysis CPUs to be provided by: Central Analysis System (CAS) at FCC: A PC/Linux dØmino back-end supplied and administrated by the computing division Remote Analysis Centers (RAC): Institutions with CPU, disk and personnel resources to serve collaborators Emphasis on Remote Analysis with formal agreements FNAL pursuing improved offsite connectivity Analysis needs must be better understood and modeled http://www-hep.uta.edu/~d0race/d0rac-wg/d0rac-final.pdf … ClueDØ … RAC (Regional Analysis Center)

6. Amber Boehnlein, FNAL Analysis Patterns Current access and analysis patterns  Physics group coordinated efforts  Derived data sets by skimming through data sets (DST or TMB)  Picked event samples of raw data for re-reco studies  Specialized reprocessing of small data sets  Physics topic analysis includes generation of test samples, trigger simulation, background studies, efficiency studies  User level analysis primarily takes place on skimmed data samples on high level tier Need to acquire more information in order to size the analysis systems and tape plant Need to simulate system to identify bottlenecks

7. Amber Boehnlein, FNAL Cost Estimate-June 2002 DØ Cost Estimate Total 200320042005200620072003-2007 Infrastructure$400,000$310,000$570,000$645,000$295,000$2,220,000 Analysis machines$1,152,500$865,000$1,152,500$1,025,000$680,000$4,875,000 FNAL CLuED0$50,000 $250,000 Reconstruction$225,000$325,000$575,000$150,000$200,000$1,475,000 Central disk cache$150,000 $200,000 $850,000 Robotic storage$75,000$0$150,000 $525,000 Tape drives$450,000 $300,000$600,000 $2,400,000 Backup facility$100,000 Sum$2,602,500$2,150,000$2,997,500$2,820,000$2,175,000$12,745,000 5 Year planning processing, with yearly review Infrastructure includes database machines, networking, web servers, code Servers and build machines

8. Amber Boehnlein, FNAL Costing Assumptions  Analysis cost was estimated by modeling possible user scenarios—Demonstrated need for collaboration resources beyond FNAL for computing, but does not reflect full system costs  Disk and servers  Networking  Infrastructure such as gateway machines and code servers  Mass storage  Cache machines and drives to support extensive data export  Project Disk not included  Disk estimate only for FNAL central SAM cache on IRIX machine  Monte Carlo estimates not included  Reprocessing not included

9. Amber Boehnlein, FNAL FNAL Equipment Budget The guidance to the experiments is $2M/year— asked to provide $1.5M scenarios for FY 2003 Reduction in farm purchase, tape drives Use the FNAL equipment budget to provide basic level of functionality  Database and other infrastructure  Reconstruction farm  Robotic storage and tape drives  Disk cache  Basic analysis computing  Support for data access to enable offsite computing

10. Amber Boehnlein, FNAL Institution Contributions All Monte Carlo production takes place at regional centers Secondary reprocessing at Michigan National Partnership for Advanced Computing Initiative Center—targeting reprocessing for 15% of data set, goal for March Additional sites needed Contributions at FNAL to project disk and to CLuED0 Offsite Analysis  Task Force prototype sites at GridKa and Lyon with routine SAM delivery of the Thumbnail data sample to GridKa  SAM-GRID prototype demonstration

11. Amber Boehnlein, FNAL Conclusions  The D0 computing model is successful SAM, an integrated data handling system, enables flexibility in the allocation of resources and effective use disk cache and robotic storage and path into the GRID era  Use FNAL Computing budget to provide base for infrastructure, robotic storage, reconstruction and analysis  Use collaborating institution resources for project disk, analysis computing, MC generation and secondary reconstruction.