Leslie Groer University of Toronto

1
ATLAS Distributed Analysis Tutorial

• Why are we here?
• Datasets are much too large to copy to local
  resources every time
• Available resources at CERN and at individual
  local sites insufficient for many tasks
• LCG model is to have distributed datasets and
  distributed resources
• Up to now the Grid has been used mostly by
  expert production users
• Goal of this workshop
• Start educating regular users how to run on
  distributed resources
• Mostly analysis jobs (AODs) and simulation
• At start up lots of ESD and even RAW data to be
  analyzed
• Make good use of Canadian resources
• Many thanks to Dietrich Liko from CERN
• Thanks to Isabel Trigger, Rod Walker, Bryan
  Caron, Rachid Mazini, Neil Knecht, Gregory Wu,
  many of you for being beta-testers

2
What is the Grid (Grid 101)
Monitoring Elements
Tape and disk storage
Centralized Services Virtual Organization
Membership Service (VOMS) Resource Brokers
(RB) LCG File Catalog (LFC) File Transfer
Services (FTS) Distributed Data Manager (DDM)
Storage Element (SE) Storage Resource Manager
(SRM)
gLite Middleware Globus toolkit GSI
Worker Nodes (WN)
User Interface (UI) Client Tools (e.g. ganga,
panda, dq2)
Compute Element (CE)
3
LHC Computing Grid (LCG)

• Canadian Tier-1
• TRIUMF
• Eastern Tier-2
• McGill
• Toronto
• Western Tier-2
• Alberta
• Simon Frasier
• Victoria
• CFI NPF resources over next 5 years

4
Canadian Resources for Tutorial

• Toronto Fish Desktop Cluster
• Outside the physics.utoronto.ca domain need to go
  through gatekeeper node guppy01.physics.utoronto.c
  a fishgate.physics.utoronto.ca
• Using 4 nodes that are running SL3.0.5, and have
  the needed setups (in particular Java, AFS)
• Use the node based on your last name
• A-F antler M-S cherub
• G-L altair T-Z conan
• Need to use port 2222 for ssh (-p2222)
• Toronto Tier-2 Bigmac HEP Cluster
• 224 compute nodes, dual Xeon 2.4 GHz (350
  kSI2k)
• 36 TB of disk (7 TB available to LCG)
• CE bigmac-lcg-ce.physics.utoronto.ca
• Running the Atlas software CVS repository
  (bigmac-m1.physics.utoronto.ca)
• TRIUMF Tier-1 Compute Cluster
• 33 compute nodes, dual and quad CPUs (250
  kSI2k)
• 24 TB disk, 12 TB tape
• CE lcgce01.triumf.ca

3.0 GHz CPU 1.3 kSI2k
5
Timetable

• 0900-0910 Introduction - Leslie Groer
  (University of Toronto)
• 0910-0940 Introduction to Ganga and Panda -
  Dietrich Liko (CERN)
• 0940-1000 Grid Computing in Canada - Rod Walker
  (TRIUMF)
• 1000-1330 Exercise 1 Distributed AOD Analysis
• (coffee available 1030, lunch available 1230)
• 1330-1350 Feedback on Exercise 1
• 1350-1650 Exercise 2 Distributed Full Chain
  (Simulation, Digitisation, Reconstruction)
• 1650-1700 Conclusion of Workshop

6

• Defined some complete datasets at TRIUMF and
  Toronto to run jobs on
• SE bigmac-dcache.physics.utoronto.ca
• SE lcgse01.triumf.ca
• Atlas-Canada VO roles are supported to get better
  priority that average Atlas user
• Tutorial Homepage
• http//www.lps.umontreal.ca/twiki/bin/view/AtlasCa
  nada/DistributedAnalysisTutorial
• Toronto setup information on the Twiki
• http//www.lps.umontreal.ca/twiki/bin/view/AtlasCa
  nada/TorontoSetup
• Wireless APs
• UTATLAS1
• UTATLAS2