RECO Status Report - September, 2002

Status of releases

• Version running on farms: p11.12.01
  • Issues / problems:
    • Bottom line tracking efficiency for Z's significantly reduced as compared to p10, due to migration away from special paths (e.g. CFT axial only).
      • p11.12.01 restores CFT axial track block in reco_analyze root tuple
      • Question: why is the efficiency what it is?
    • d0_analyze crashing (10% of files).
      • New farm script does a retry, with a good success rate.
      • This executable is being phased out for production in p13.
    • The thumbnail requires db access to analyze.
      • Fix coded and scheduled for release in t02.35.00
    • Offline calorimeter geometry not aligned with trackers
      • Causes problems with algorithms that extrapolate tracks to calorimeter
      • semreco and caltrack_extra will be modified to have a temporary fix
      • The calorimeter_geometry will be updated for p13
    • L3PhysicsResults seems to be missing from L3Chunk - under investigation.

• Next pass release: p11.13.00
  • Modify semreco and caltrack_extra to handle the fact that the offline calorimeter geometry is not aligned with the tracking detectors. An ad hoc correction will be made in these packages to take care of the global offset.

• Latest production release: p12.03.00
  • Status - not certified for production use
  • Plans: This production release is being used to debug the thumbnail. There are currently no plans to install this version of RECO on the farms for data production.
  • Recent tests indicate reasonable performance. This version may be useful for Monte Carlo production.
    • Issue: get correct calorimeter and muon geometries implemented in MC and supported in RECO (under study).

• Test release
  • RECO (as of t02.35.00)
    • Average statistics concerning tracks, muons, electrons, MET are similar to p11
    • Reprocessing of DST tested.
    • Clean-up of framework rcp's underway.
  • DST / TMB
    • DST size: ~ 210 KB
    • TMB size: ~ 8 KB (data)
    • TMB+ size: ~ 21 KB (data)
  • RECO_ANALYZE - being phased out. Not supported for production use.
  • Current outstanding problems / issues (t02.35.00)
    • RECO crashes in cal_util::CorrEemcalo::corretada_IC_S (fix submitted)
    • L3PhysicsResults not saved correctly (under investigation)
    • d0om packing thumbnail errors
      • Fix for emid submitted
    • Turning off special stream
  • Recent performance statistics
    • t02.32.00

• Punch list for p13
  • Test new calorimeter geometry - data & MC
    • data version being tested
  • Test new muon geometry - MC
  • Install default tracking algorithm
    • Use new GtrMetaPkg to control algorithms
  • Support axial-only tracks in 401 chunk
  • Check setup for measuring tracking efficiencies using MC hits
  • Double-check reprocessing
    • p13 DST (data, MC)
    • p11 DST (data, MC)
  • Check TMB content
  • Review error messages
  • Fix L3PhysicsResults

Recent progress:

• DST size
  • GTrack chunk size reduced x2
• TMB size has reached its goal.
• Reprocessing of DST works.
• Memory requirements have been reduced significantly.

Major outstanding problems / issues:

• We have not yet deployed access to the detector calibration databases.
  • Issues:
    • Memory requirements - d0omCORBA is being improve
    • Server reliability - not yet tested
• There is a worry that analyses requiring a measurement of MET will require the CalDataChunk. This implies continuing with the "TMB+" in p13. An estimate was made that this would require an addition 3 TB of disk space if we processed data at 25 Hz through February, 2003. The Top and WZ groups have made explicit requests for this functionality. Current plan - support TMB+ until 02/03.
• With the current track finding inefficiency, the identification of photons is compromised. To improve this situation, CFT axial-only tracks can be used. Current plan - try to include these tracks in the 401 chunk, and assign their eta values to something unphysical in chpart_reco. This will allow these tracks to be included in the thumbnail, but easily identifiable for analyses. Status: being deployed.
• There is a critical need to develop an Event based Error chunk, that can keep track of things like fatal unpacking errors. This information also needs to be propogated to the TMB. This is one of the projects previously advertised.

Ongoing projects:

• Tracking algorithms for p13 - A task force collected data concerning the currently available tracking algorithms and recommended that the combined GTR/HTF algorithms should be the default. Effort is now required to implement this recommendation and to confirm that it is installed properly for p13.
• Calorimeter - A task force is studying the issue of the calorimeter zero-suppression threshold (online, Level 3 and offline).
• Effort reports from all Algorithm / Object ID groups being organized
• p13 goals from all Algorithm / Object ID groups being organized
• A cleanup of the error logger messages from RECO is ongoing.
• Investigation into physical dependences of packages has started, with the goal of reducing the memory required to link an analysis executable.
• Discussions to make more direct connection between TMB and TMB Tree objects is underway. Proposal to ID / Physics groups has been sent.
• Currently analyzing a DST or TMB requires access to the run configuration database. This is fixed in t02.35.00.
• A new RECO certification system has been developed that will improve testing / certification / monitoring of results. Version one of this system has been deployed in the test release.
• RECO Monitoring tasks - Currently nine people have volunteered and are working on the two advertised RECO monitoring projects, however the level of effort has to date been rather low. There are many additional aspects of these projects that need volunteers.
• Chunk documentation - Currently three people have volunteered and others are participating, however the convergence rate is rather slow. More people are needed.

This page maintained by Harry Melanson