Run Quality Coordination

1
Run Quality Coordination
Run Quality Coordination Meeting Feb 12th, 2003

• The charge
• Basic Ideas
• What we can do now and later

2
Draft Charge

• This charge still being thought out by the great
  powers. I wrote
• Our physics interests require emphasis on
  attaining and maintaining the highest standard of
  data quality.
• Many separate efforts going on at
  Expert/Online/Offline/Physics Gp levels
• We are charged to
• Increase communication between efforts
• Provide feedback to the appropriate detector and
  software groups
• Provide feedforward to the analysis and physics
  groups
• Identify places where we need effort.

3
Summary of D0 Data Quality Discussion

• Special GCAS Meeting Jan 27th
• Agenda was
• Identify what we are doing well
• Identify where we need an improvement and hear
  new, untried ideas
• Identify people who are willing to help
• Jianming said he was going to put the minutes
  on the web.

4
What is working well

• Lots of tools and different levels of monitoring
• The polish and importance of the alarms is
  increasing
• Experienced daq-shifters paying off
• Global monitoring is beginning to make a
  contribution

5
Where we need improvement

• Lots of tools and error messages but often we
  dont follow up. Shifters are taught to ignore
  error messages
• Example the unpacker shows errors. The expert
  says he/she is aware of them but cannot fix them
  at the moment
• The readout monitor said almost all the PDTs had
  readout errors but the expert said they were
  expected and could be ignored with this version
• Shifters are urged to log and report all problems

6
New Ideas

• Various Levels of data quality control
• The run Run Quality Database
• The luminosity block GS implementing idea to
  overcome the latency loophole in the alarm
  system -gt thumbnails? An alarm database (browser
  tools for the SES log)?
• The Event Muon Raw data quality -gt thumbnail
• The Reco version Reco quality assessment for
  every run evaluated by the Offline shifter
• Software alarms connected to Coor

7
Unused Assets

• Offline shifters to look at every run
• Requires a way to glom-together some output from
  all parts of a run (Sangjoon Lee) to make
• Standard plots and histograms
• Need someplace to check off that things are
  good/bad (all work should have a product).
• Call chain in case of problems

8
What I think we need to do next

• Gather complete list of what data quality control
  is going on in each system and by whom
• Understand what is the best technical solution to
  the Reco-data-glomming problem
• What do we look at? Tuples? Is so, where/how much
  data is cached?
• What do we look at? Thumbnails? If so, what
  histos and which package?
• Increase the usefulness of Global Monitoring
• Advocate more standard plots from the experts
• Advocate increased communication to experts
• Questions or Serious problems -gt phone call list
• Less serious problems -gt email list
• Carefully get complete, 4 p coverage into the
  Run Quality Database (remember 0.9525 0).

9
Points of Monitoring Quality Control
Good Runs List Detector

• Could be

Recorded Data Online
Alarms SES
Data- Taking
Reco Farms
Examine
Recorded Data Offline
Analysis Samples
Event Check And Marking
Good Runs List Reco Version Dep.
10
Summary

• We need to have the best quality data so that we
  get the best quality results.
• There are diverse and serious efforts
• I dont know how this coordination effort is
  going to be run
• I am making it up as I go along.
• Do we have meetings?

11
We want to avoid this
Headline on D0 World News

• Perfectly good muon data wrecked!
• Usual suspects rounded up for questioning.
• Its happened two or three times to us.
• It could happen to you.
• It might be preventable.

• I wrote this slide for the Oklahoma workshop.
• I dont recall what the problem was that time but
  I bet it was more subtle.
• I used it again in December