The UK Grid Report

1
The UK Grid Report

• Paul Jeffreys
• Rutherford Appleton Laboratory and PPARC
• Historical context
• Particle Physics initiatives in the UK
• UK Particle Physics computational grid
  initiatives
• Grid co-ordination across the UK
• UK Particle Physics grid organisation
• (Networking activities)
• Summary

2
Historical Perspective

• 1Q 1999
• Cashmore Panel looks into UK PP programme
  beyond 2005 recognised that substantial
  resources needed for computing
• Based loosely on grid model
• Bid under preparation for LHC computing
  infrastructure in UK (to Joint Infrastructure
  Fund) but unable to reach agreement on details
• 3Q 1999
• Government announces that Spending Review
  brought forward by one year, submit Feb/Mar 2000,
  announcement Autumn 2000 -- for 2001-3
• JIF LHC bid steered towards case for prototype
  grid in the UK
• Funding Council (PPARC Jim Sadlier
  representative at workshop) recognises
  opportunity to bid against SR for computational
  grid, as cross-disciplinary exercise
• Resources required for Particle Physics grid
  defined - based on analysis made for JIF bid
• 4Q 1999
• Join CERN co-ordinated European Grid project
  Crucially important as impossible for UK to
  operate independently
• Every institute in UK signs up for JIF bid and
  submitted (vitally important as demonstrates
  whole community behind initiative)

3
Particle Physics Initiatives

• BaBar experiment bid to Joint Research Equipment
  Initiative fund
• Prepared 1998, early form of grid model
• Awarded 800k (Jan99), of which 650k funded
  kit worth 2M
• Forced to go out for re-tender SUN won
• Equipment for 10 UK sites, RAL is Regional
  Centre
• One of three Regional Centres world-wide
• Mutual support between centres
• RAL received
• E4500 processor (6 CPUs), 4GB memory and 4.5TB
  (out of 12TB total) disk
• Staff post to develop central database system
  (just started)
• Universities have not had effort to develop local
  facilities yet
• Effort is again real issue
• LHCb JREI bid for MAP
• Liverpool bid
• Awarded 270k (Jan99), less than requested
• Purchase 300 PCs to operate LINUX farm
• 107 events simulated per week just keep
  Ntuples
• No opportunity for full interactive analysis
• Running since November 1999

4
Particle Physics Initiatives

• CDF/D0 experiments at Fermilab (1999)
• CDF - successful Joint Infrastructure Fund bid
• Submitted Dec98, successful in Dec99
• FNAL for UK physicists 10TB disk, 4SMP
  workstations
• RAL 5TB disk, 5TB tape, SMP workstation
• (possibly split to be reconsidered)
• (Some provision for MINOS)
• 4Mbps transatlantic connection first funded
  connection!
• Little at 4 universities
• D0 - successful JREI bid .. Regional Centre at
  Lancaster
• Lancaster bid (access to Manchester/IC) submitted
  May 99
• COMPAQ industrial partner
• 60Gips of PC power, 24TB Exabyte(!) store (5X
  less than disk), 1TB
• 1 FTE for 2 years
• All facilities at Lancaster, none at FNAL, tape
  not disk!
• Likely to be successful

5
Particle Physic Initiatives

• BaBar - JIF
• Submitted April 99, total value .. 3M
• Line to SLAC at 4Mbps ..1.2M
• Computers to exploit equipment bought under
  JREI
• Big SMP at RAL (E6500 -8 cpu) .. 0.3M
• Six LINUX simulation farms at Birmingham,
  Bristol, Manchester, QMW, Brunel and RHBNC ..
  0.4M
• Held over hear spring 2000
• Summary
• Real funds for particle physics computing (new
  pots of money!)
• No national co-ordination possible (in fact
  opposite encouraged!)
• Good experience already in regional centre for
  BaBar, Liverpool MAP, and planning for other
  centres
• UK Particle Physics grid offers opportunity to
  give some coherence across national computing
  provision
• Same comment as Richard M ? vital that not
  perceived as isolated activities to win funds
  although clear overhead in co-ordinating

6
LHC JIF bid for prototype grid

• LHC experiment bid to JIF
• Prototype Tier 1 centre, located at Rutherford
  Appleton Laboratory
• Agreed by community that prototype Tier 1 should
  be at RAL
• Will be notified in Autumn 2000 (possibly
  indication earlier)
• Funds available across 2001,2,3
• Total bid for 6M
• 3100 PC99, 125 TB disk, 0.3PB tape, 50Mbps
  network to Tier 0
• 10 staff at centre, 10 based in universities
  (staff crucial component)
• Staff need to be sold as software!!
• Prototype Tier 2 centres (Edinburgh, Liverpool)
• Look for close industrial participation
• (Collaboration with other scientific disciplines
  discussed below)

7
Submission to be made to S.R.

• UK aiming for single Tier 1 centre to support all
  4 LHC experiments
• Analysis indicates community not large enough to
  support gt1
• Tier 1 assumed to support 100 physicists in 4
  working groups
• Start-up
• Processors 14400 PC99 equivalents
• Disk 296 TB
• Tape 2.0 PB
• Networking (Tier 0) Gbps
• Effort Increasing to 35SY (10 of in
  universities)
• 9SY experiment specific software and
  applications
• 9SY middleware
• 17SY lower level, networking, 24x7 operation
• Not hard-wired, we recognise that middleware is
  the major challenge
• Crucial Point no new funds for grid
  specifically until April 2001

8
Cross-disciplinary Developments

• Recognised widely (OST) -- national computational
  grid developments must be science driven -
  E-science
• Hence SR bid initiated and co-ordinated by
  Research Councils
• SR bid titled..
• Information Technology and the UK Science and
  Engineering Base
• Meeting held 17 January essentially to develop
  research IT green-paper
• Particle Physics
• BioScience
• Environmental
• Physical Sciences
• Very marked contrast between PP and rest. We have
  a well specified problem to a well defined
  timescale, which has to be addressed, and
  coherent community
• Almost embarrassingly good response for PP!
• Meeting encourages
• PP activity on grid as presented - sees PP grid
  as prototype for others (avoids question of
  combining and diffusing activities)
• Bio-science community
• Environmental community ... but recognised more
  diffuse (in UK)

9
Cross-disciplinary Developments

• Propose meetings between infrastructure providers
  (CLRC/UKERNA/HPC) centres and scientific
  disciplines
• PWJ asked to co-ordinate PPA with
  infrastructure communities
• Combined with BioScience 13 March 2000
• Computer Scientists and Industry invited
• Key gathering
• Membership of Wellcome IT Management Team
• PPARC Organising Trip for Companies interested in
  Grid to CERN
• 9-10 March
• Jim Sadlier
• Links being made through CLRC
• For example .. Computing Science and Engineering
  group have links with Bio-Informatics people who
  are keen to try out some applications on a grid
• Benefit of multi-disciplinary institute

10
Particle Physics Grid Co-ordination

• PP A met on 24 January to prepare initial work
  schedule
• Desire to integrate Astronomy side (as
  appropriate)
• Key aspect of future plans are to work closely
  with CERN and Data-Intensive Computational Grid
  team and others (GLOBUS?)
• Participate in testbed activities (EU grid
  project)
• Identify areas in which UK can work which
  complements other activities
• Evaluate middleware and look to contribute
• Investigate QoS and other network issues
• Next months will be used to define action plan
  (as shown for PPDG)
• Half Day UK Town Meeting 15 March 2000
• Meeting for UK LHC collaborations and community
• Key-note talk from Richard Mount
• Objective inform PP (and A) community of all
  developments, and in particular seek their
  support
• Recruit people to work on Grid from universities
• Important as rapid developments

11
Particle Physics Grid Co-ordination

• CLRC Grid Team formed
• Commissioned 27 January
• Within Particle Physics, squeezed out from
  existing activities
• IT Department .. Re-direct some effort paid for
  by PP community
• Computational Science and Engineering support
  (area of common interest)
• Looking for (3-5 2-3 1-2) SY (intended to be
  at least 7SY)
• Hoping to recruit effort need some people
  working most of their time
• Team will build up and naturally move into JIF
  era (to level of 1010 staff?)
• UK grid management team being formed
• Consist of representatives from Tier 2 possible
  centres, and sites representing successful
  computing bids, plus CLRC, plus collaboration
  representatives,
• Oversee activities on national level
• Given that substantial resources are likely to
  become available, crucial that managed correctly,
  transparently, coherently

12
Networking

• UK has had an active interest in networking for
  many years
• Recently worked closely with Les Cottrell on ICFA
  monitoring
• Sophisticated network monitoring since c.1996
  (used by UKERNA!)
• Number of interesting networking developments and
  tests
• need to commission 4Mbps RAL to Chicago
  (non-trivial as transatlantic connection
  terminates in London)
• hopefully will have funds for 50Mbps connection
  to CERN
• Tests
• QoS tests from US to UK
• Presently set up from Abilene to JANET (CAR,
  WRED) .. Only incoming
• But we need to get into Abilene (Les looking into
  this)
• Will also have ESnet to JANET priority (been
  promised since last summer)
• Managed Bandwidth to CERN
• Commissioned UKERNA to set up PVC over SuperJanet
  and Ten-155
• Should be set up by the summer
• Discover that Japanese link to CERN via London
  with MB!!
• Managed bandwidth to BaBar university sites from
  RAL (JCG)
• PVC over ATM, but may be limited to SuperJANET3

13
Summary

• Exciting time in the UK
• New pots of money, and potential funds for grid
  from 2001
• Support in high places for grid development
• Particle Physics encouraged to blaze the trail
• Very keen to do this, but in close collaboration
  with rest of PP community
• Responsibility to encourage other disciplines in
  process
• Intend to establish close partnership with
  industry from the beginning
• Seen at many levels as
• A great opportunity .. -- and a threat!

14
PPARC .. Overview of Near-future Activities

PPARC-Collaboration Link HI-phi -- Meeting FORUM,
more required?
IT Focus for UK R.Community
Links to US PPDG (talk) Globus
IT kit from bids BaBar CDF,...
CERN Co-ordination Framework V
Computer Scientists Tony Hey
LHC Comp. Challenge JIF???
PPARC PLANNING
JT/OST Link SR2000 Work Groups
Industry
Other Disciplines
Tier2 centres Edin, Lpool, ...
LHC Computing Review
UKERNA JIF bid QoS tests to US SJ-4 GEANT
MONARC Contribute
Egrid Other inits
Bio-informatics