1. Level-1 Calo Monitoring
Level-1 Architecture
Data sources and monitoring rationale
Monitoring tasks
Detailed rate histograms
Internal consistency
Calorimeter comparison
CTP comparison
Module status
Archiving
Stephen Hillier, University of Birmingham
July 7th 2005

2. Level-1 Architecture and Monitoring data sources
Preprocessor
124 modules
Cluster
Processor
56 modules
Jet/Energy
32 modules
Merging
8 modules
4 modules
Digitized
Energies
Calorimeter
Signals
Merged
Results
To CTP
Real-time
Data Path
Readout
Driver (ROD)
14 modules
Data
Region of
Interest ROD
6 modules
Interest Data
Stephen Hillier, University of Birmingham
July 7th 2005

3. Level-1 Architecture and Monitoring data sources
Preprocessor
124 modules
Cluster
Processor
56 modules
Jet/Energy
32 modules
Merging
8 modules
4 modules
Digitized
Energies
Calorimeter
Signals
Merged
Results
To CTP
Real-time
Data Path
Readout
Driver (ROD)
14 modules
Data
Region of
Interest ROD
6 modules
Interest Data
4. CTP
comparison
2. Internal
consistency
3. Calorimeter
1. Rate
histograms
5. Module Status
Stephen Hillier, University of Birmingham
July 7th 2005

4. Detailed Rate Histograms
PPM sees ‘raw’ trigger towers
Hardware histogram capability
Rates
Hit maps
BC Number dependence
By far the largest monitoring data source (bar event stream)
Analysis route
VME read to SBC
Publish to ‘gatherer’ in local PC?
Export upstairs at regular intervals?
Update time O(1 minute)
Monitoring
Farm?
Surface
USA15 x8
VME read
Stephen Hillier, University of Birmingham
July 7th 2005

5. Stephen Hillier, University of Birmingham
Internal Consistency
L1Calo data is routed via 6 ROSes
3 PPM
1 CP system
1 JEP system
1 RoI data
Full internal consistency checks require event building
SFI, EF, SFO
Event rate
No hard requirement
As many as possible
Results
Hopefully very small – error flags, location of errors, MRS messages?
Ideally one flag saying NO PROBLEMS!
Some data monitoring performed at ROS
CP, JEP, RoI hit maps (no need for gatherer)
Relatively small amount of data
Stephen Hillier, University of Birmingham
July 7th 2005

6. Calorimeter Comparison
Requirements similar to internal consistency
Needs Liquid Argon and Tile data
Therefore requires event building
Event rate: as many as possible
Results: a number (?) of histograms, errors, warnings
Probably smallish (but not as small as digital comparisons)
Calibration Notes
We do not intend to use physics runs for online calibration
Calibration requires dedicated runs
Which probably will use monitoring tools, but not at high rate
And probably eventual offline fine calibration
But monitoring will be used to verify current calibrations
eg reference histogram checking for correlations
Stephen Hillier, University of Birmingham
July 7th 2005

7. Stephen Hillier, University of Birmingham
CTP Comparison
Requirements similar again
Needs CTP data
But data size is trivial
L1Calo sends ~100 bits to CTP
But important they are correct!
Stephen Hillier, University of Birmingham
July 7th 2005

8. Stephen Hillier, University of Birmingham
Module Status
Relatively small number of parameters read by VME from every module
Link status, parity errors, overflows, logic faults
Not (in general) DCS type information
Typically <100 words per module
But remember system has ~300 modules
Error flags should be alerted immediately!
Publish frequency ‘high’ O(1 second)
Need good GUI in control room
Immediate alerts
Quick fault identification
Stephen Hillier, University of Birmingham
July 7th 2005

9. Stephen Hillier, University of Birmingham
Data Archiving
Ideal:
save all data every run
keep everything forever
and immediately accesible
More realistic proposal:
Run summary histograms
at most 10 histos, keep forever, quick access
Detailed summary histograms
O(100) histos, keep forever, slow access
All data and histograms
Keep O(1 week), quick access
Stephen Hillier, University of Birmingham
July 7th 2005