Physics Data Processing at NIKHEF

1
Physics Data Processing at NIKHEF

• Jeff Templon

WAR 7 May 2004
2
Goals

1. Realize an LHC physics computing infrastructure
   optimized for use by NIKHEF physicists
2. Where possible, combine expertise associated with
   goal 1 for other projects with NIKHEF
   participation
3. Capitalize on expertise available funds by
   participating in closely-related EU NL projects
4. Use NIKHEF Grid-computing expertise and capacity
   as currency

3
NIKHEF-optimal LHC Computing Infrastructure

• Operation of LCG core site
• Build experience with site operation and discover
  external site issues traditionally ignored by
  CERN
• Leverage front-runner position earned by our EDG
  effort
• Strong participation in LHC/LCG/HEP Grid
  framework projects
• Meten is weten preoptimization is the root of
  all evil (Knuth)
• Leverage front-runner position earned by EDG
  effort
• Leading role in Architecture/Design arm of EGEE
• AA model is fulcrum of balance between
  CERN-centric and really distributed models
• Make use of accumulated expertise in security
  to gain position in middleware design
• Preparation for Tier-1
• Avoids having others determine NIKHEF computing
  priorities

4
LHC/LCG/HEP Projects

• Strong Coupling to NIKHEF LHC experiment analysis
• One grad student per experiment, working with
  ARDA project early influence, experience, and
  expertise with LHC analysis frameworks
• Room for more participation in medium term
  (postdocs, staff)
• Continuing work with D0 reprocessing
• D0 metadata model is far advanced compared to LHC
  model
• Influence via our (LHC) task-distribution
  expertise on US computing
• Investigations on ATLAS distributed Level-3
  trigger
• Precursor for LOFAR/Km3NeT activities

5
Preparation for Tier-1

• Tier-1 for LHC
• Archive 1/7 of raw data, all ESDs produced on
  site, all MC produced on site, full copies of AOD
  and tags
• Contribute 1/7 of twice-yearly reprocessing
  power
• End result major computing facility in
  Watergraafsmeer
• 1 petabyte each of disk cache tape store per
  year start 2008
• 2000 CPUs in 2008
• 1.5 Gbit/s network to CERN
• These numbers are per experiment
• NIKHEF contributes research, SARA eventually
  takes lions share of operation
• NCF must underwrite this effort (MoU with CERN)

6
Overlap with other NIKHEF projects

• Other HEP experiments
• D0 work Q4 2003, continuing
• Babar project together with Miron Livny
  (Wisconsin)
• Astroparticle physics
• LOFAR SOC much like LHC Tier-1
• Km3NeT on-demand repointing much like ATLAS
  Level-3 trigger

7
EU NL Projects

• EGEE (EU FP6 project, 22 years, 30M)
• Funding for site operation (together with SARA)
• Funding for Grid Technology projects (together
  with UvA)
• Funding for generic applications (read non-LHC)
• BSIK/VL-E
• Funding for Data-Intensive Science (everything we
  do)
• Funding for Scaling and Validation (large-scale
  site operation)
• Cooperation with other disciplines
• Leverage multi-disciplinary use of our
  infrastructure into large NCF-funded facility
  (Tier-1)

8
Currency

• Advantages of Grid Computing for external funding
• Grid computing (cycles expertise) in exchange
  for membership fees

9
People

• LHC applications
• Templon, Bos, postdoc, 3 grad students
• Non-LHC applications
• Van Leeuwen (CT), Grijpink (CT), Bos, Templon,
  Groep
• Grid Technology
• Groep, Koeroo (CT), Venekamp (CT), Steenbakkers
  (UvA), Templon
• Site Operations
• Salomoni (CT), Groep, Templon, other CT support

10
People / Funding

• EGEE
• 1 FTE Generic Apps, 1 FTE Site Operations, 1 FTE
  AA
• BSIK/VL-E
• 1 FTE Scaling Validation, 1 FTE Data-Intensive
  Sciences
• Both projects require local 1-1 matching (50
  cost model)
• Can overlap - 15
• Possible additional money from bio-range project
• Possible to replace some manpower with equivalent
  equipment

11
Possible Funding Model