Participating in Globus Test-bed Activity for D

1
Participating in Globus Test-bed Activity for
DØGrid

• UTA HEP group is playing a leading role in
  establishing DØGrid and remote computing systems,
  in addition to ATLAS
• Since the DØ experiment is taking data currently,
  this is the best platform to test out Grid
  computing architecture for LHC experiment such as
  ATLAS
• Playing leading role in the DØ Grid and its
  test-bed activity and linking it to ATLAS will
  increase UTAs chance to host an ATLAS Tier II
  computing site
• An MRI proposal to acquire sufficient hardware
  for a DØ Regional Analysis Center (DØRAC) has
  been submitted ? UTA is the only US candidate at
  the moment for DØRAC
• UTA-HEP group has been participating in various
  Grid Test-bed activities through ATLAS and DØ
  Experiments and independently
• The more activity we contribute the higher the
  chance for Tier II

2
Globus Remote Job Submission Test

• Globus is one of the available Grid Job
  submission and security tool kit but not a batch
  system
• Globus is being explored for adoption for DØ
• Globus can and must work with many different
  batch control systems PBS, Condor, Condor-G, LSF
• HEP group has been working with ACS toward
  utilizing the ample Intel Linux cpu power for DØ
  detector simulation, but due to its LSF batch
  control system the process has been slow
• This works in our advantage for Globus remote job
  submission testing for LSF system

3
How does Globus Remote Job Submission work?

• Globus Job Submission Sequence
• An authorized user with a proper certificate
  submits a job to the gatekeeper
• The Gatekeeper checks the certificate with the
  central authority
• The central authority gives back the key for
  clearance
• A Globus account in Gatekeeper submits the job to
  the batch controller
• Batch controller assigns a batch queue for the
  processing
• Batch controller tells back to the Gatekeeper for
  completion
• Gatekeeper sends results to the user

2
1
3
8
UTA-ACS
4
What do we need from ACS?

• An ACS Intel machine as the Globus gatekeeper
• A Globus account (normal user privilege)
• Install Globus took kit must be installed on the
  gatekeeper as root (Need help from ACS personnel)
• Work with the ACS contact to get Globus set up,
  including certification application and other
  configuration
• Batch job will be a simple non-cpu critical
  Hello world! style.
• Testing will be sporadic and brief.
• Total Testing Duration Not clearly known yet but
  presume at least a couple of months
• Contact persons
• HEP Tomasz Wlodek Jae Yu (x2814)
• FNAL Gabrielle Garzoglio (garzogli_at_fnal.gov)

5
Some detailed technical information

• ACS Linux farm runs RH6.2 with a planned upgrade
  to RH7.2 sometime in the summer.  Does this
  matter for the testing?
• No.
• The globus toolkit 2.0 is distributed for Linux
  2.x/Intel x86. The same software should work for
  both the version minor configuration changes.
• ACS's LSF system is version 4.1.  Does this
  matter?
• Probably not due to minimal use of LSF command in
  Globus
• This is part of the testing
• How much disk space is the testing going to
  require?
• Globus toolkit 150 MB.
• Log files 15 KB/Mo.
• No plan for implementing large executable in the
  first phase
• Globus system stability
• No system crash in 3 months testing of globus
• However, no experience with LSF yet

6

• How much of security risk is this testing going
  to be?
• Very minimal
• Globus Security Infrastructure (GSI) uses
  asymmetric (PKI) mechanisms to enforce security
  which is used in e-business web sites for
  transactions
• Number of people participating in testing
• Currently 5 non-UTA participants
• Expect to grow at a minimal level to less than 20
• UTA HEP group is a strong participant
• Not all of them will run on UTA ACS system
• What will this lead to?  Is this just purely a
  one-time test, or will it lead to significant use
  of UTA's supercomputer system by others (non-UTA
  people)?
• Two phases of testing is planned
• Phase 1 Minimal Hello World! style testing
  that uses extremely insignificant resources
• Phase 2 More complicated DØ experiment
  executable with several months after phase1
• Resource usage even for Phase 2 is not expected
  to large or extensive
• This testing will likely to continue as long as
  it is allowed to go on

7
Proposed Operating Principles for Globus Testing

• ACS Administrator has a complete control over the
  testing executable
• At any time the ACS administrators notice issues,
  the testing will shutdown till a resolution
• If the testing seriously interferes with ACS LSF
  operation, the testing should stop and a
  resolution needs to be found
• The contact persons at UTA and Fermilab will be
  responsible for regulating and coordinating the
  flow of testing