Choose timezone
Your profile timezone:
JCOP Project Team Meeting
There were approximately 24 people at this meeting.
1) Minutes
There were no comments to the minutes of the last meeting and so these were approved.
2) News
Sascha Schmeling noted that he had received very few responses to the questionnaire
regarding training for PVSS and the FW. He requested that people in the experiments
be encouraged to complete this questionnaire as it will help in the process of
defining the extension of the current training to cover the evolving needs of the
experiments. New courses are being prepared and the feedback from the questionnaire
will be useful in this process.
The questionnaire can be found under:
http://cern.ch/pvssregister/JCOPCourseQuestionnaire.asp
3) Summary of the Result of Phase 2 of the PVSS Scaling Up Project
Paul Burkimsher presented the results of the second phase of the Scaling Up
Project. His slides can be found under:
http://indico.cern.ch/conferenceDisplay.py?confId=a057390
The following questions/comments were raised:
- Clara Gaspar asked whether the test in which the Event Manager (EV)
was 'scattered' onto a different machine from the rest of the system had been done
on Windows, Linux or both. Paul replied that this had been done on Windows. Clara
noted that the results would probably be significantly different for Linux.
- Lennart Jirdén asked whether the Alert Screen (AES) automatically restarted once
an avalanche had finished (i.e. once the overload had finished). Paul responded
that in the case of a temporary pause that this was indeed the case. However, if
the AES had been stopped then the operator would have to click on the appropriate
button to restart it.
- David Myers asked whether the EV was 'scattered' on a dual processor machine.
Paul noted that there were performance improvements when running on a dual
processor machine but that these were not as significant as on two machines since
disk access became a limiting factor before CPU load. The OS automatically
distributes the manager across the two processors but it is also possible to
manually allocate a manager to one or other processor.
- Clara noted that disk access limitations are often reached due to the last value
archive. However, if this is switched off, then archiving is also disabled. ETM
have been requested to change this behaviour.
- On the point of whether having many alarms configured had an impact on
performance or not, Jim Cook asked whether the performance impact only came on an
alarm state change or also if the value remained in the alarm state. Paul responded
that the performance hit was related only to the change of state (CAME/WENT) and
not with the fact that the dpe is in alarm.
- David noted that information presented by Paul or discussed in the FAQs should
be disseminated to the maximum possible extent in the experiments.
- Paul encouraged people to use the controls Wiki that he has set up. David
enquired whether it was possible to control the content and also to see who
submitted something. Paul replied that he controls the content and that it was
possible to see who submitted a particular entry.
- Clara asked whether the items not yet covered will be looked at. Paul responded
that he has other things to look at and one would have to look at relative
priorities. The experiments should perhaps re-review what has already been done and
what they would like to be done in addition.
4) A.o.B.
Wayne reminded the meeting that there would be a meeting with ETM on the 19th/20th December to discuss CERN's priorities for PVSS 3.5. A pre-meeting will be held on the 14th and everyone should submit their priorities to Wayne prior to this emeting.
There are minutes attached to this event.
Show them.