Line: 1 to 1 | ||||||||
---|---|---|---|---|---|---|---|---|
Minutes of T2UK meeting at RHUL 6th Feb 2008Present: Fred, Gordon, Nikos, Pedro, Andrzej,Barry, XXX, Jiri, Paul, | ||||||||
Changed: | ||||||||
< < | John, Simon, YYY, Bill, Stefan Ask, PPPP | |||||||
> > | John, Simon, Tracy, Bill, Stefan, YYY, Thorsten | |||||||
Minutes last meeting: No comments. | ||||||||
Line: 36 to 36 | ||||||||
ROSFred reported that the CERN group have been carrying out some ROS recabling. Technical run ongoing this week checking out tdaq-01-08-04. | ||||||||
Changed: | ||||||||
< < | Nobody asked about ROS software. | |||||||
> > | ||||||||
HLT farmsFred reported that our american colleagues have to spend money now. He's been helping them to choose which machines to buy. ~500 machines to arrive in CERN around end Feb/ begining of March for use in April. More performant quad cores than | ||||||||
Changed: | ||||||||
< < | existing XPUs. | |||||||
> > | existing CPUs. | |||||||
NetworksSecond controls core switch installed and running. Second core data | ||||||||
Line: 54 to 54 | ||||||||
CSC note progressing (new draft scheduled for Friday)
Jiri | ||||||||
Changed: | ||||||||
< < | EF tracking performance, less fake tracks than L2... | |||||||
> > | EF tracking performance, timing etc. Less fake tracks than L2, details on the slides. | |||||||
Some discussion on clarity of plot where fake tracks beyond range of | ||||||||
Changed: | ||||||||
< < | generated momentum. should be clarified in text of CSC note | |||||||
> > | generated momentum. should be clarified in text of CSC note. | |||||||
StefanID DQMF segment included in November tech run. Only have few automatic tests (keep it manageble). Using test for "not empty" histos can see which algs inactive (grey), OK (green) and error (red). | ||||||||
Changed: | ||||||||
< < | List of quantities to monitor discussed with tire-0 monitoring people, | |||||||
> > | List of quantities to monitor discussed with tier-0 monitoring people, | |||||||
initial list created. | ||||||||
Line: 73 to 74 | ||||||||
No big surprises. Data preparation timing proportional to RoI size.
John | ||||||||
Changed: | ||||||||
< < | Optimisation for different RoI types. IDscan need to look at B- | |||||||
> > | Optimisation for different RoI types. IDscan need to look at optimising the tau, B-phys and bjet slice. | |||||||
Look at more realistic and problematic data samples. Corrupting (truncating) fragments in bytestream files. At L2 ID code runs without crashing. Dmitry to change code to return error code if | ||||||||
Line: 86 to 88 | ||||||||
TrigIndetTrack biggest problem (runs both SiTrack and IDSCAN +...) having multiple copies of tracks due to the way persistency works (Dmitry has halved the size taken by tracks in | ||||||||
Changed: | ||||||||
< < | persistency). | |||||||
> > | persistency). Need to discuss/decide what amount of info goes into the different data formats (ESD/AOD/DPD) | |||||||
NikosShould think about how to run Fex algorithms from ESD | ||||||||
Line: 119 to 122 | ||||||||
DQMF or straioght comparison of histogram files.
Region selector: | ||||||||
Changed: | ||||||||
< < | New region selector ...use for M6? | |||||||
> > | New region selector to be pushed into the nightlies asap, must have it by for rel-14. | |||||||
Online spacepoint tool should be changed to use new region selector, 4fold increase in performance. Still need to make changes to enable/disable ROBs from OKS dtabase. Nikos to forward mail about new interface to Dmitry and Jiri. | ||||||||
Changed: | ||||||||
< < | Mark to look at TRT code as well | |||||||
> > | Mark to look at TRT code as well. | |||||||
HLT steeringNothing much to say, most of the work already coverd by EDM point | ||||||||
Line: 158 to 162 | ||||||||
Any other items people want should be requested ASAP. Travel 49K spent (40K LTA 9K short visit). Still enough left for more short trips this year. | ||||||||
Changed: | ||||||||
< < | ||||||||
> > | Claims to go in asap by the end of March. | |||||||
Line: 1 to 1 | ||||||||
---|---|---|---|---|---|---|---|---|
Added: | ||||||||
> > |
Minutes of T2UK meeting at RHUL 6th Feb 2008Present: Fred, Gordon, Nikos, Pedro, Andrzej,Barry, XXX, Jiri, Paul, John, Simon, YYY, Bill, Stefan Ask, PPPP Minutes last meeting: No comments.NewsFDR underway. Shipping data to tier-0 and runnig reconstruction (which is currently crashing). Starting with nominal geometry, calculated alignment constants (1.5.days on a single PC). ATLAS detector paper has been submitted (around Xmas). CSC notes coming (March). Schedule not officially changed (rumours of 1 month delay, mid-July). Cooling of sectors 3-4 weeks behind schedule, machine checkout & hardware commissioning now overlapping.Recent ActivitiesRobinsA couple of new problems since last meeting. One has shown the symptoms of bad PLX chip (some of the latest German production have similar problem so we're not alone). Would like a "robinExpress" same as robin but with a PCI express interface instead of PCI-X. Should start buying components before the money evaporates. Discussion of robin replacement for SLHC. Discussions next week in CERN with Philippe to gather requirements.ROSFred reported that the CERN group have been carrying out some ROS recabling. Technical run ongoing this week checking out tdaq-01-08-04. Nobody asked about ROS software.HLT farmsFred reported that our american colleagues have to spend money now. He's been helping them to choose which machines to buy. ~500 machines to arrive in CERN around end Feb/ begining of March for use in April. More performant quad cores than existing XPUs.NetworksSecond controls core switch installed and running. Second core data switch installed. Currently some ROS data goes through both switches, 2nd stage is to make everything dual pathed.HLT trackingCSC note progressing (new draft scheduled for Friday)JiriEF tracking performance, less fake tracks than L2... Some discussion on clarity of plot where fake tracks beyond range of generated momentum. should be clarified in text of CSC noteStefanID DQMF segment included in November tech run. Only have few automatic tests (keep it manageble). Using test for "not empty" histos can see which algs inactive (grey), OK (green) and error (red). List of quantities to monitor discussed with tire-0 monitoring people, initial list created.DmitrySlides on web page. Looking at severl slices. No big surprises. Data preparation timing proportional to RoI size.JohnOptimisation for different RoI types. IDscan need to look at B- Look at more realistic and problematic data samples. Corrupting (truncating) fragments in bytestream files. At L2 ID code runs without crashing. Dmitry to change code to return error code if unpacking has a problem. Hope to have some guidelines by robustness workshop (4th March). EF use converters which don't have a return code so must use exceptions, follow up with offline people. John would like to investigate effects status codes in ROB header.EDM meetingTrigIndetTrack biggest problem (runs both SiTrack and IDSCAN +...) having multiple copies of tracks due to the way persistency works (Dmitry has halved the size taken by tracks in persistency).NikosShould think about how to run Fex algorithms from ESD Simon: Data model group told to put in document that trigger people will always need access to raw data.MonitoringDuring M5 used different version of PT for Atlantis montoring which can run asynchronously with the TDAQ run. Working on tdaq-01-08-04 version for M6.Release coordination (Simon)13.0.40.2 Was to be used for FDR but not ready yet so they're using a nightly 13.2.0 for M6 got ready by Ricardo 14.0.0 Due end of Feb 14.1, 14.2 ... every 6 weeks Simon coordinating 14.0 for trigger, need more coordinatorsValidation (Ricardo)Moving ref files from each test to afs 13.2.0 afew non-critical patches in P1 cache discussion of the difference between DQM and software validation. Same quantities histogrammed but different tools used for checking DQMF or straioght comparison of histogram files.Region selector:New region selector ...use for M6? Online spacepoint tool should be changed to use new region selector, 4fold increase in performance. Still need to make changes to enable/disable ROBs from OKS dtabase. Nikos to forward mail about new interface to Dmitry and Jiri. Mark to look at TRT code as wellHLT steeringNothing much to say, most of the work already coverd by EDM point above and Trigger config next.Trigger Configuration(Paul):slides on twiki. Documentation "not ideal but growing". John: Should there be another way to change histogram ranges? Keep separate so monitoring desk can change tem but should not have write access to trigger database. More generally, discussion of having different roles for using trigger tool with different access permissions. Discussion of applying a change to multiple objects (e.g. changing a prescale for many items gives you RSI).Menus3 Bphys CSC notes merging to oneNext meeting
FinanceCurrent spend 65K another 90K just being signed off. Need to order compnents for robinExpress before end of year Any other items people want should be requested ASAP. Travel 49K spent (40K LTA 9K short visit). Still enough left for more short trips this year. -- GordonCrone - 07 Feb 2008 |