JEB Todo List
The list of items to address and implement for the JEB system prior to completion.
- "Plan A" release to Pole (Late September)
- Implementing the same filtering as Plan B.
- Move JEB/trunk development to new "V2" icetray offline software (mid October)
- Make use of new "interfaces" project, include stuff from JEB in interfaces
- Reorganize JEB projects somewhat, make dependencies clean
- Evaluate performance impacts of v2 goodies
- Lazy frame stuff
- I3Outlet (Nov/Dec?)
- cleanup code
- use lazy frame stuff
- Handle GCD frames as well
- Will allow server to make single DB query and distribute to clients
- Include in offline once finished?
- jebserver module performance (Nov/Dec?)
- input/output queues to have independent mutexes.
- PDAQ interfaces (Gent meeting)
- Trigger enumerations in daq-decode (persuade DAQ group to never reuse integers!)
- Status/monitoring
- Finalize list of items to check and report to ExpCont? for Operator monitoring.
- Collect system monitoring information by using subcomponents, the component control and the remote control. Possible subcomponents are
- DAQDispatchReaderService (file name, events and bytes read from file, event rate, byte rate)
- I3Outlet (name, #connected queues, #mandatory queues, #missing mandatory queues plus names)
- I3InletService (name, status of connection)
- JEBEventService (run, subrun and event ID, merging mode, error messages, last flush, last autoflush)
- JEBServer (#clients, size, #idle frames, #not client filtered frames)
- I3PFFilterModule (filter mode, rate to pot and crop, rates of all filter streams, trigger rate)
- I3JEBWriter (file name)
- tba (run, subrun and event ID, frame rate, started at, running since)
- Reduce full system monitoring information within jebsystem script and report a subset to EC only.
- Analysis and monitoring clients
- Would include online CnV, realtime data analysis clients (TOO, GRB/burst notification, event displays etc)
- Anything that need access to all events post reconstruction and filtering
- Idea to reuse the jebserver module only this time to post filtered frames.
- In std jebserver (could block jebserver from processing based filtering)
- Separate post-filtering, separate server
- Would pick up files from local disk cache from PFRaw hardlink directory.
- Would include online CnV, realtime data analysis clients (TOO, GRB/burst notification, event displays etc)
- Persistency needed?
- Likely depends on frequency and causes of crashes
- Dispatches? Or just read entire file and if a few seconds of data is lost at crash time, C'est la vie.
- JEB server system: Dump all frame contents to disk at "crash" time, and reload at startup?
- Modes of operation
- Filtering modes: PhysicsFiltering?, RandomFiltering?, NoFiltAllToTape?, NoFiltAllToSat?
- Merging modes: MergeEvents?, I3DAQOnly, TWRDAQOnly, ReadBothNoMerge?
- Are these lists complete?
- Non-standard filtering
- Use case: Filtering of calibration-type data
- Possible: always run calibration filters in std filtering suite and control filter on/off per run?
- Possible: Branch to a second server (talking to second set of Calibration filter clients) based on run type.
- Watchdog.
- Possible: Each component has a built in "no activity" watchdog which kills hanging processes
- Possible: Single, central watchdog. Uses control infrastructure to kill/restart hung processes
- Serve double duty as central monitoring collection point?
- Check TWR keys while decoding file/system header
- Documentation
- Better end-user documentation with:
- General starting/stopping instructions
- Subsystem overview
- Troubleshooting information
- Better end-user documentation with:
Last modified 10 years ago
Last modified on Apr 28, 2014 2:55:41 PM