Project

General

Profile

2016-01-08-SwC

Software & Computing Issues

Software & Computing Issues:

  • Efficiencies
    • Tracker sign flip & geometry fixed for real data. Still appears to be a flip in the MC. Fixes are currently being validated.
    • With real data, cluster & space-point efficiencies seem OK, but track-finding efficiency is low (~75%), and worse for SSD.
      • Need to trace it through pattern recognition. AD is heading this.
    • TOF0 space-point efficiency is low ~70% (expect a slightly lower efficiency for TOF0 compared to TOF1,2. DR has started looking into it, some evidence that the efficiency is lower for the horizontal slabs.
      • Possible that it will have to be recalibrated (in software)
  • Data reprocessing
    • Need a consistent official reconstructed data-set with correct geometries & all bug fixes.
    • DR has been processing data on MLCR with official/patch releases of MAUS.
    • Expect a new mainline release (and possibly new geometry) after the current round of tracker MC validations.
    • Will then reprocess all Step IV data with that release.
    • Longer term question of whether to hand off reprocessing to the GRID (so as to not clog e.g during data-taking) or continue on MLCR offrec machine.
  • MC production
    • Batch MC production needs to be started up.
    • There are some structural issues with handling the G4BL inputs on the GRID. Was discussed at CM42 and handed to the Serbian group. This activity needs to be picked up urgently. Also need a long-term MC production manager to push jobs through & monitor on the GRID
    • In the meanwhile can do MC production on the MLCR machines (DR)
    • Speed remains an issue that needs to be sorted. The slowness seems to come from G4, and there's some evidence that it's slower for field-on than -off. Speed investigation is currently resource limited.
  • Network
    • It was agreed that the network interface to C&M needs to be made more robust and isolated from any potential glitches.
    • Plan is to
      • use statically assigned IP addresses
      • avoid software name lookup
      • lockout incoming ssh connections to C&M machines during magnet commissioning & production running
    • Requires
      • PMH to go through C&M address list and reference hardware by IP (estimate from PMH is week of Jan 13, but then requires testing failure modes.
      • PMH to provide a PV to indicate running mode
      • PF to provide a script to lockout ssh based on running mode