Minutes for 14 June 2006

ATLAS detector paper

Fred reported that Dave Francis and Thorsten Wengler have been nominated to be editors for the TDAQ part of the ATLAS detector paper. Thorsten pointed out that TDAQ will have only about 15 pages max. in this paper, and that the main work will be to agree on a set of supporting papers and make sure they get published. Pedro pointed out that papers guaranteed to be finished until the end of the year 2007 will be considered in the RAE, and that this statement has to be made for each paper by October 2007.

ROBINs

Production of the ROBINs has finished and the boards have been extensively tested at CERN. 8 boards still have problems that need to be investigated. The German batch shows a similar rate of failures. 20 more cards will be produced in Germany, to cater for late-comers like FP420, and for additional spares.

ROS & HLT PCs

The message was received that Elonex has gone bankrupt, and the question was raised what the impact on the ROS PC delivery and maintenance is. It turned out during the meeting that only the part of Elonex dealing with digital home kits was in trouble, and that the PC delivery should not be affected. The danger of being left without support when the vendor goes bust was recognised, but is seen as an unavoidable risk when dealing with commercial companies, that can only be minimised by carefully choosing the vendor, but not completely avoided.

The newly announced 'Woodcrest' Xeon processors attracted a lot of attention. Intel tests show they achieve over 120 on SPECint. A brief discussion was held on using dual socket, dual core processors running multi-threaded code to process four events in parallel, or alternatively four instances of the same code. It was also noted that it may be worth waiting for AMD's response.

Severe performance hits have been noted with the 2.4 kernel and dual core processors. Need to finalise the specification for Nov 2006 machines very soon, which could possibly be these new Intel ones (expected to be released in June).

Network Implementation

Richard Hughes-Jones reported that the network plans contained only small accounting errors and that the order for chassis has been put in. The GPS on the roof of the Manchester building has been re-connected.

For the medium scale test we require of order 200-300 machines. We need to decide which site to use in the next two weeks, which can be either West Grid (Canada), Manchester or LxShare. Alessandra Forti is running the Manchester Tier 2 effort. West Grid and LxShare will make the machines available by taking them off the grid, while Manchester will run the jobs inside the grid. The first has the advantage that the environment is closer to the real system, the latter that the system can be available not only for a single shot test, but at short notice and over a longer period of time.

Richard reported that the packets and bytes in the switches are monitored by a commercial system.

Finances, LTA's etc:

At the moment we have one LTA, Tania. Thorsten Wengler - HLT Royal Holloway - no update UCL - 1 from January

AOB: Technical Papers

Pedro raised the need for ideas for possible papers to include in the ATLAS technical paper. The deadline is October 2007 to announce the paper. Thorsten Wengler on the editorial board. Referenced papers need to available by October next year. This will be an item of the next face-to-face T2UK meeting.

Next Meetings

The next phone meeting will he held on Thursday 31 August at 11.00, organised by T. Wengler. The next face-to-face meeting will be held on Wednesday 11 Oct 2006, possibly at Royal Holloway.

Work Package 3

T. Wengler shows slides giving an Overview of Trigger Configuration. The need for the Trigger DB to be available at Grid sites was emphasised. Also, a Java front-end for the Trigger DB, known as the Trigger Tool was discussed. At the moment it can be used by "Trigger Experts" but the need for non-experts to be able to use this was noted.

Also discussed was how to distribute the load between the 3000 nodes x cores x processors. A large scale test will be performed at SLAC for scalability.

Integration of HLT, Interface between Trigger DB and (new) steering being developed for 12.x.x (not 12.0.0). This should be done by the end of summer.

Started a new software project: DetectorCommon. Simon should say more.

Milestones: July : LVL1 sim + hardware, HLT steering, JobOption service Sept : Production system ready for use

Nikos: What is the effort on the UK side on configuration? Thorsten: Tania on TriggerTool for HLT. James Keates has 1 -> 1.5 months, could start working on monitoring issues for HLT algorithms, to be discussed. Simon Head - leading in from top physics activity. Paul Bell in autumn (after SCT commitments) Un-ki Yang (new lecturer Manchester) - Next Month moves to Manchester, has expressed interest in trigger work Christian Schwanenberger (new lecturer Manchester) from September, has expressed interest in trigger work

Steering and trigger tool essential (3-6 months)

HLT software

The HLT software should be tested in one to two more weeks. At the moment it requires 3 to 4 experts to use, get this down to one, and then one non-expert. Region selector is now a service in Athena releases 12.0.1 and 12.1.0. The Bug at high eta, in earlier versions, is now gone. Only tested the ID. Action UCL to extend to muons.

Releases and Automated Tests

Athena release 12.0.1 is scheduled for the 20 June, and the last in branch for the end of July. We have to have everything in place by then. Release 12.1.0 will use gcc 3.4.4 as default, which causes some packages to fail build now. These are being fixed and the fixes are also being putting in to 12.0.1.

A brief discussion on which platform should be used for the software tests was held, either SLC3 or 4. Ask Simon.

Try to automate tests, such as the e/gamma slice test. Action Simon.

LVL2 Tracking

Slides were presented on the zFinder improvement studies performed by Nikos, Mark and Erkcan. The Efficiency was low for low pT particles, with zFinder currently the weakest link. The study was performed with Athena 11.0.3 on Muons in the barrel. New alternatives for zFinder yield a better resolution, as an improvement for single particles. However, the errors on dz and dr are larger than expected. This needs following up.

John Baines presented slides. Most objects are persistifiable. Currently working on LVL2 association between tracks and generator particles. First version in a release soon.

Muon Reconstruction

Lily. Slides shown for a study on 6 GeV muons using mufast. Needs more investigation, with higher energy muons.

E gamma slice tests

John Baines. Done several fixes and multi-threaded mode. Still work to do.

Monitoring

Nikos. We require Histogramming inside algorithms. Action James Keates.

Cosmics

Bypass region selector and it works in the offline software. See above for monitoring.

ID scan

Nikos. Promised to do documentation for various parts of ID scan.

Physics Performance

Monika raised questions about the plots required for monitoring. Problems in the EF ID tracking were noted. The next trigger performance week is next week.

Validation for Athena release 12 is due to start soon, in which the CBNT has been removed and replaced by AAN. This requires an update to the code, which is easy but dull. For start up we need trigger menus and to decide what to do on day one. Comments were made that we are not making the EF work enough.

Low Energy Electrons

Bill. Low efficiency for 5 GeV electrons. Large improvements in 11.0.4 and 11.0.5 over 10.0.1. Still need to look at pileup.

B Physics

John Baines commented that we now have a complete trigger slice for the EF in 12.0.x. Also, an Atlas note of Julie's RoI B studies is being created. This uses fairly large jet RoIs. This study should be repeated with release 12. Need to improve efficiency.

-- NikosKonstantinidis - 14 Aug 2006