the science of simulation

1
the science of simulation
computer architectures
non-perturbative QFT
machines
algorithms
phenomenology
falsification
better theories
experimental tests
2
UKQCD, the first UK Grand Challenge

• commercial computers
• optimised codes ? competitive edge
• effort focused on exploitation

3
has it been a success?

• yes excellent value for money
• machines, codes and data have been fully
  exploited
• but we had to concentrate on one action and one
  parameter range
• zero temperature O(a)-improved Wilson quarks
• and we were dependent on purchasing machines off
  the shelf
• first the UK parallel computing industry
  disappeared
• now the HPC market is shrinking world wide

4
UKQCD international partnerships

• DOEPPARCRIKEN QCDOC
• low power ASIC using IBM embedded processor
  technology to scale to 104 nodes
• DOE SciDAC
• standard APIs for portability made a prerequisite
  for large machine funding
• linked to UKQCD via QCDOC open source policy
• PPARC GridPP
• QCD data grid conforming to emerging EDG OGSA
  standards
• linked to SciDAC via QCDOC

5
the UK QCDgrid
Tier 2
Edinburgh
Glasgow
Liverpool
QCDOC
Tier 1
Swansea
Tier 0
? US DOE SciDAC
6
the next step

• there is a robust software strategy
• open source codes will provide choice
• the grid will provide tools for data integration
• OGSA standard will allow interoperability
• but who will build and operate machines?
• we are dependent on a few individuals and face
  rapidly changing market forces

the challenge

• to encourage machine builders
• and make codes data available to all

7
International Particle Simulation Organisation

• Purpose
• to generate and manage field theory
  configurations
• Funding
• annual subscriptions based on community size and
  value of total data held
• subscriptions paid with data and/or money
• data value decreases according to Moores law
• Council
• decides actions and parameters
• manages the grid
• buys data, providing funding for machines
• Users
• from member states may access all data