Present: Lily, Nikos, Erkcan, Mark, Thorsten, Pedro, Ricardo, Simon, Bill, Fred, Monika

ACTION: Fred to circulate a note with details of the bid for new build machines to the group leaders.

News

Nikos summarised the report from the ATLAS-Uk OsC. It is generally positive, and they are particularly happy with the progress in the ROBins and the fact that the few failures are not due to a single design fault. They asked three specific things for the next meeting (which is on 03/11/2006 at CERN).

• More details on the plan for acquisition of CPUs (and back-up plan in case the CPUs do not meet the specs we have been assuming in our calculations)
• A definition of the T2UK project completion
• Progress on the T2UK deliverables and how they fit into the overall ATLAS TDAQ plan

RobIns

Fred: The bulk production has been done, of the 350 produced in the UK, 7 do not work. Of these 2 may be recovered, 5 seem lost. For the German batch of the same size, 10 seem dead. A further production will be launched in Germany, to accommodate needs of new clients, such as the forward detectors, and make sure there are enough spares.

The installation is progressing well, with the majority of RoBin already installed, and many already commissioned standalone. Commissioning with the respective RODs in some cases needs to wait for the availability of the RODs.

ROS s/w

Gordon is away at the moment, Fred reported that much work is still being done on extra bits of functionality and fixes fro tdaq-1.6. Further changes are foreseen to be necessary for tadq-1.7.

Further HLT h/w

Fred reported that the purchasing of processing nodes is somewhat delayed, partly because not all funding agencies have signed the pledge letters yet, partly because of developments on the hardware side. Intel has introduced a new series of chips with very promising specs. First tests show that the increase in performance is not as much as the specs suggest, but these machines are the fastest ones tried so far. The plan is to have about 100 nodes by the end of the year, which will be used for commissioning.

Networking

Richard was not present. Fred reported that the infrastructure installation is going ahead at speed, all the pizza boxes for 2006 and the chassis are available, installation is being carried out by the network group. The bill for the UK contribution has not arrived yet, but is expected soon.

Release news

Simon reported that there is much work going currently into validating and bug fixing code for 12.0.3, which is expected to be built on Tuesday, Sep. 5th. After that, there will be a release 12.0.4. Release 12.0.3 will be used for simulation, then for determining calibration parameters, 12.0.4 will then be used with the new parameters for reconstruction. 12.0.4 will be too late for the LST - even 12.0.3 looks tight, so there will be a special release for the LST based on 12.0.3.

Steering

Much development going on in the development branch, new steering expected to be used in production with rel 13. The development process appear to be on track, with some shortcomings in documenting changes, which caused difficulties for some developers.

Configuration

Thorsten reported that the work on the trigger configuration is going well, with the TrigConfig review coming up during September as a major milestone. Much work goes into getting the software and documentation ready for the review. The full software chain is in place, but a complete cycle of uploading a running setup to the database, and then running it from the database with identical results is still to be done (hopefully in the next few days). Tania is made much progress on the HLT part of the TriggerTool (GUI to the trigger database), and will also work on the java version of the upload scripts used to insert running setups into the database. The configuration group also takes an active interest in the LST, where a running setup from the database will be required.

Medium scale tests (MST)

Thorsten reported on the preparation of the medium scale tests to be carried out on the Manchester grid farm. People involved are Alessandra Forti, Richard Hughes-Jones, Thorsten Wengler in Manchester, the HLT team running the tests will be Sarah Wheeler, Hegoi Garitaonandia, Haimo Zobernig, with help from Marc Dobson, and other people from the TDAQ community as appropriate. Recently, Sarah Demers from SLAC has joined the discussions with the aim to test the DBProxy in the MST. The test will likely be carried out in two parts: using tdaq-1.4. which allows the use of algorithms, and the newer tdaq-1.6. as the latest release which will be used in the LST. The MST will aim to scale the system up to about 300-400 machines, using the GRID nodes (in GRID mode with full accounting) in Manchester. Once working, this should give a good environment for future tests on large numbers of machines without lengthy negotiations to reserve a large cluster somewhere.

Further s/w news

There were no news on the region selectior. Simon reported that he has made progress on the automatic testing, which still needs some further work before it will be fully automatic, but progress made so far already makes it easier to run the tests.

L2 Tracking and Physics performance

Mark and Erkcan reported on improvements made on the Z-finder (see slides on Wiki page). An effort has been made to improve the efficiency in RoIs with single electrons or muons of low pT (6/10 GeV) by using a new formula to determine how many neigbors in phi should be considered in each layer. The new code shows a marked improvements on the RoIs lost so far and also improves the overall resolution, reducing significantly the tails of the distribution.

Nikos reported on progress he and Julie made in tuning the pattern recognition for low momentum tracks on L2, for Bphysics and b-jet tagging. The rate of events that are reconstructed at the Event but fail IDscan is now down to about 1%. Monika reported that a lot of effort goes into validating the e/g slice s/w for 12.0.3 with most of the attention focussing on the EF. More details will be presented at todays PESA meeting.

Thorsten reported that James Keates has been working with John and others to produce monitoring algorithms for IDscan. Partly this s/w is already committed, partly it will wait until after release 12.0.3., where it represents larger changes. James has now finished his period with full time involvement with HLT monitoring and will concentrate on the theory part of his PhD from now on, but will still be available on the ~10% level for updates and adding of features to the code he has written.

Riccardo reported on work he has done on a trigger decision object, which allows the user to access information on which trigger signatures were run in an event, and which passed. He will also adopt this to the new steering. He has also been producing new and simpler configuration files for Z->ee studies on L2.

Finances

Fred reported that some spending of CORE money was under way, but well within expectations. On the travel budget about 4 k have been used, which is less thanexpected, but not overly so. People travelling on appropriate business are encouraged to use budget FK039. The first bills for LTA have started to come through as well. On new bids the group producing the HLT images have asked for faster machines - this is seem to be a reasonable request, and it is expected that 5-7k will be spend soon on buying some faster machines for this purpose.

AOB

Next meeting will be a face-to-face meeting at Royal Holloway, on the 11th of October, starting around lunch time.

Fred encouraged people to properly register themselves in their offices at CERN and with the correct percentage in the users office, to support tdaq's requests for office space.

-- NikosKonstantinidis - 04 Sep 2006