DAQ 100 - (Jeff Landgraf)

1
DAQ 100 - (Jeff Landgraf)

• Only write clusters to .daq files
• Remove some current bottlenecks in DAQ
• - Cluster finder optimization
• - Formatting optimization / bypass
• - Formatting buffer management

This Gives us

• 5-fold reduction in data size
• High instantaneous rates approaching
• 100 Hz (Min Bias 50MB/sec)
• 50 Hz (Central 100MB/sec)
• Better dead time management
• Last year 20 Hz MinBias ? 80-100 dead
• This year 20 Hz MinBias ? 20 dead
• Full buffered use of the 30 MB/sec (2.7 TB/day)
  HPSS link
• Could approach 1.3 million Central/day
• 5 million MinBias/day

Last years record 2.0 TB 210,000
Central events
2

• STATUS
• DAQ technical bottlenecks
• Cluster finder optimization (done)
• Formatting optimization / bypass (done)
• Formatting / cluster buffer management (done)
• Second myrinet card on BB (done)
• Study Disks on BB (done / not implemented)
• Clustering / interface issues
• Cluster finder ? Offline Chain (done)
• Cluster finder physics evaluation
  (ongoing)
• Gain Corrections and T0 corrections need to be
  done online
• Mechanism inside DAQ exists
• Need procedures for the corrections
  (august)
• (Automated manual)
• Need procedures for monitoring them
• (Automated in EVP Shift requirements)
• Offline chain needs to be aware of correction
  status

3
Comparison between DAQ / TCL
cluster finders
/t3 (daq) /t4 (tcl)
nEvents 9123 9123
nReconstructed 9011 9120
nVtx 5243 5310
nGlobal 6.28E6 6.46E6
nPrimary 2.54E6 2.54E6
4
DAQ cluster finder gets

• More events with large number of primary tracks
• Fewer events with large number of global tracks

5
DAQ Cluster finder gets

• Smaller RMS distance from global track to hits
• More short tracks (more track splitting)
• Better peak for long tracks

Why? Consistent with wide distribution of
tracking cuts in tcl
6
2-Hit Resolution (Fabrice Retiere)
7
Track Splitting (Mercedes Noriega)
8
Track Merging (Mercedes Noriega)
9
Correlation Functions (Mercedes Noriega)
3D correlation functions
10

• Summary
• We should be able to turn on clusters for next
  run.
• Initial studies on the quality of the clusters
  are promising
• Responsibility for evaluation of the cluster
  finder should
• be (and is) moving to the physics groups