GlueX Offline Meeting, December 21, 2016

From GlueXWiki
Revision as of 10:37, 4 January 2017 by Marki (Talk | contribs) (Upcoming OSG Service at JLab)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

GlueX Offline Software Meeting
Wednesday, December 21, 2016
2:00 pm EST
JLab: CEBAF Center F326/327

Agenda

  1. Announcements
    1. CentOS 7 upgrades; farm outages for relocation Dec 19-20
    2. Problems with reading recent data with many threads (David)
  2. Review of minutes from the last meeting (all)
  3. Report from the JLab Computing Steering Committee Meeting (Mark)
  4. Launches (Paul M./Alex A.)
  5. Sim 1.2 (Sean, Mark)
  6. HDGeant4 Update (Richard)
  7. Upcoming OSG Service at JLab (Richard)
  8. Review of recent pull requests (all)
  9. Review of recent discussion on the Gluex Software Help List.
  10. Action Item Review

Communication Information

Remote Connection

Slides

Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2016 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2016/ .

Minutes

Present:

  • CMU: Naomi Jarvis, Curtis Meyer
  • JLab: Alexander Austregesilo, Thomas Britton, Brad Cannon, Sean Dobbs, Sergey Furletov, Mark Ito (chair), Mahmoud Kamel
  • MIT: Cristiano Fanelli
  • UConn: Richard Jones
  • W&M: Justin Stevens

There is a recording of this meeting on the BlueJeans site.

Announcements

  1. CentOS 7 upgrades; farm outages for relocation Dec 19-20. Conversion of farm nodes to CentOS7 continues. Monday and Tuesday several nodes were physically relocated in the Data Center.
  2. Problems with reading recent data with many threads. This problem appears to be due to corruption in the raw data. The cause is under investigation.

Review of minutes from the last meeting

We went over the minutes from December 7.

  • We will have to have a discussion about whether HDDS XML files will continue to be present in the Git repository once they are served from the CCDB. If they remain, it might be confusing, but it will be what we are used to.
  • Sean pointed out that work with the new scheme for access to HDDS information should be coordinated with work on the new event viewer.

Report from the JLab Computing Steering Committee Meeting

Mark gave the report. This is a standing meeting which is called every few months. It is the forum where IT Division reports progress to all of the various computing stakeholders around the lab. Mark flashed slides from CNI and SciComp to illustrate the scope of the subjects covered (slides are password protected, username starts with "g", ends in "x", password that of the our document portal thing).

Launches

Alex gave the report. He had remarks beyond those in his email from earlier in the week.

  • There has been a large reduction in the memory needed to run the analysis launch since it got started. There were several changes that Paul Mattione made to the analysis libraries to do this. In the beginning, jobs were run using only 18 threads of the 72 hyperthreads available on the farm nodes. Now all 72 can be used, although the job speed only increased by a a factor of two. Now the jobs use 50 GB of the 64 GB available.
  • He would like to start the "real" monitoring launch before the break. Sean mentioned that there are still some calibrations that need to get finished before the launch can start, namely timing in the tagger, PS, and TOF.
  • The wiki page describing the second analysis launch on the Spring data is in preparation. There are a lot more channels to catalog this time. There was an effort to eliminate plugins for channels that were made redundant by the Mitchell Omni-plugin, but there still may be some reduction possible.

Sim 1.2

Sean is almost ready with the necessary mcsmear parameters. He is looking at the following systems:

  • Start Counter. He is looking at whether tracking uncertainties are adding a significant amount to the measured resolution. If so those effects should not be attributed to the intrinsic start counter resolution.
  • FDC. He is tuning some of the thresholds being applied for the FDC cathodes.
  • CDC. Some checks on the efficiency plug-in are planned.

We talked about generating electromagnetic background for the simulated events. Right now this is dominating the CPU time need. Sean has plans for overlaying the hits from real data on the simulated hits.

HDGeant4 Update

Richard gave us a report.

  • He has added the geometry for the DIRC and the triplet polarimeter and is working on hit generation for the DIRC now.
  • The boost libraries are required to build HDGeant4. These are available as optional packages in most Linux distributions.
  • Event generation for triplet events will be done inside of HDGeant4 itself and not in an external generator program. If this functionality is required, you need to build Richard's DiracXX package, available from GitHub (it is one of his public repositories).
  • Mark and Nathan Sparks are working on automated builds.
  • If folks find problems or have suggestions, they should open an issue on GitHub.

Upcoming OSG Service at JLab

Richard announced a new facility at JLab for job submission to the Open Science Grid (OSG). There will be a new box installed at the lab, spec'ed out by the OSG folks, with the latest OSG software stack installed and an robust network connection to the outside world. The box will be administered by SciComp. Users can log in with their CUE credentials. To submit jobs from the GlueX Virtual Organization, one needs a personal OSG credential, available through Richard. See his wiki page on OSG use for instructions on how to obtain one.

This is a pilot project at present. If it proves to be successful and easy to use it could give a big boost in opportunistic computing power widely available to all collaborators.