January 14th, Softlab, bldg 864
Participants: Jorg Wenninger, Ralph Steinhagen, Kris Kostro, Jens Andersson (summary)
Orbit control prototyping will continue in 2004. New orbit server version planned.
Topics:
- Timing
Perhaps the biggest question mark. Should move to a timing-driven approach. Which event should signify that a new calculation should be done with currently available data, and how should it be delivered? Can the energy (and other machine state) be delivered the same way? How often is it necessary to sync with others? Should investigate what BDI and PO are using.- Hardware
No immediate need for hardware (except for possibly the timing), except for a dedicated machine to continue SPS prototyping (abcmw1 requested). The multiprocessor Proliants could be used as well, but no dedicated machine required.- General design
Post mortem and state read-out; not block real-time data flow. Sketch below.- Cycles
The 'cycle' concept will disappear for the LHC version. There will be different modes instead, perhaps with an external program changing config at certain times. For SPS it can be seen as a restart each cycle.- Puzzling and data logging
Puzzling late data together should not be the work of the orbit server. Some data logging can perhaps be done by an external user.
To note: Jens is leaving after the summer -- plans to do a new version of the orbit server now, to be tested and used during 2004 and possible to reuse in the future.
Sketch: PDF version ; Visio version
More info added afterwards
January 15th, lunch with Lars Jensen:
At the moment, and also expected in the future, BDI's acquisition is triggered only by turn count, and not synchronized to GPS. There is a discussion about locking BST to UTC, but currently no plans or any BDI need to do so.