Particle Physics Data Grid (PPDG) project

1
Particle Physics Data Grid (PPDG) project

• Les Cottrell SLAC
• Presented at the NGI workshop, Berkeley, 7/21/99

2
Who are we?

• Collaboration of 9 sites
• 6 DoE Labs ANL, BNL, FNAL, JLAB, LBNL, SLAC
• 3 universities CalTech, U. Wisconsin, SDSC
• Collaboration of
• particle physicists
• with long history of large collaborations
• actively promote WANs
• pioneers in use of networks/collaborations/large
  data sets
• computer scientists
• researching and developing tools infrastructure
  

3
Project Objectives

• Delivery of infrastructure for
• widely distributed analysis of particle physics
  data
• at multi petabyte scales
• by thousands of physicists
• Acceleration of development of
• network middleware infrastructure aimed at data
  intensive collaboration science

4
Problem

• Analysis of Particle Physics data
• Scale of accelerators detectors means only a
  few sites
• So large collaboration (hundreds to thousands) of
  physicists from tens to hundreds of sites in tens
  of countries
• Create large data sets with hundreds of Terabytes
  to Petabytes of data over the next decade
• Some sites already taking data (Jlab, SLAC), some
  will turn on soon (FNAL), LHC at CERN data taking
  in 2005, creating simulation data today

5
Goal relation to NGI

• Goal (impossible dream)
• researchers at universities all over US could
  query retrieve hundreds of Terabytes within
  seconds
• More than a bandwidth problem
• requires creation of intelligent networks
  network aware applications
• Stimulates NGI in 3 areas
• DiffServ to allow bulk data to co-exist
• Distributed caching
• Robustness of distributed services

6
Contributions

• ANL Globus grid middleware services
• SLAC Objectivity Open File System
• Caltech Globally Interconnected Object Databases
  project
• FNAL A data access framework (Sequential Access
  Method)
• LBNL Storage Access Coordination System
• Wisconsin Condor distributed resource management
  system
• SDSC Storage Resource Broker (SRM)
• 2Pbytes robotic storage (_at_ 6 sites), 1300
  Giga-ops/sec (8 sites), 40TB disk cache (7
  sites)

7
Applications (physics data sources)

• BNL RHIC data
• FNAL RunII/D0 CDF
• SLAC PEP-II/BaBar
• Caltech, ANL LHC/Atlas CMS
• JLab Nuclear physics experiments
• Needs
• bulk transfer of raw data to and reconstructed
  data between regional data centers - today use
  tapes, introduces delays of 2 weeks
• code builds at multiple sites
• interactive access to analyzed events
• many hundreds of Mbps growing to several Gbps in
  next few years