Notes from HPS DAQ/Offline Review


data catalog
project management updates to timeline
conditions system
readiness stance
mock data challenge

offline, Matt:

daq Sergey B.:

sep oct
install oct nov
beam mid nov
next year spring, real data

Ryan svt:

how is pilelined read-out handled



GlueX Meeting Report, March 20, 2013

FIU contract

  • Werner submitted technical proposal to University, circulated among committee members.


  • first-article counters constructed, 2 regular, 2 short, 2 narrow
  • mini-TOF has been put together, some minor mechanical adjustment issues finessed, looks pretty good
  • readiness review on Friday,
    • Beni, Chuck, Bruce will travel.


  • Meeting this afternoon, first since collaboration meeting.
  • HDDM changes, rationalizing method for storing information known only by the simulation, clear separation, pre-cursor to Geant 4 conversion.

Ideas for a data challenge

  • create a committee to oversee the effort
  • appoint a chair to the committee
  • set an unambitious scope with the understanding that scope will probably expand
  • as a suggestion for initial scope, create a system where:
    • simple (“one command”) launching of Pythia simulation jobs on JLab batch farm
      • automatic archiving of results in case of success
      • capture of job status (at least final status, in-progress status if possible)
    • simple launching of reconstruction jobs
      • user-friendly identitification of simulation data runs to be reconstructed
      • automatic archive of results
      • capture of job status

Questions from committee, software review


All Halls:

What # times to reprocess was assumed in the storage and CPU estimates?

Discuss your tools for software quality assurance.

Hall B/D:

Describe your system for cataloging and managing the data (including data quality)?

Hall D:

At what level of parallelism will serial l/O R/W hit a limit?

What is the planned latency between data acquired and completed calibrated reconstruction?

Hall B:

Provide a written outline of a plan for data challenges including scaling up end-to-end processing to a reasonable fraction of production running.

What information is included in the large amount of reconstructed data? Why do you need the reconstructed data for the full sample?