Minutes of the T2UK face-to-face meeting

UCL, Tuesday 9th Oct., 2007

(Meeting organized by RHUL; met at UCL)

Present: Manchester: Stefan Ask, Paul Bell, Simon Head, Jiri Masik. RAL: John Baines, Dmitry Emelyianov, Fred Wickens, Monika Wielers. RHUL: Antonella De Santo, Simon George, Barry Green, Pedro Teixeira-Dias (minutes). UCL: Mario Campanelli, Cathrin Bernius, Gordon Crone, Nikos Konstantinidis, Erkcan Ozcan, Mark Sutton.

Apologies: None.

News

LHC schedule: there was a presentation by Lyn Evans today. He reported on delays due to the Liquid He transport and storage, as well as on the "plug-in modules". At the moment no changes to the schedule (first collisions in July 2008) are foreseen.

The new TDAQ structure (TDMT: Livio Mapelli and Chris Bee; and Nick Ellis as overall Trigger project coordinator) has now been operational for several months. It seems to be working well.

Work packages - progress over the last 6 months and forward look

WP1: ROS hardware and software

ROBins

Fred, Barry

2 boards had soldering issues. 8 problem boards so far. Repairs attempted on all 8 (2 last week only). 3 out of 6 with PLX repair are now fully working and operational at CERN. Out of the 8 boards whose PLX were removed, 3 are now OK, possibly 4. Problem not getting any worse, no new failures of the same type recently.

Reuse of old PLX/FPGA: not clear this will work; need to re-ball the Ball Grid Array, and CemGraft has counselled against this.

Final German production (additional 70 boards) was finished end of August.

Next TDAQ release tdaq-01-09 will incorporate a CRC checksum that will be attempted on a fraction of the data, to detect any data corruption problems in the ROBin.

ROS software

Gordon

tdaq-01-08.03 now installed in P1. Will be used in the next commissioning run. Changed the way information is reported to IS, towards OMD (Operational Monitoring Display). tdaq-01-08.03 also includes the CRC checksum.

ROS operational monitoring

e.g., identify hot ROSs; OMD functionality still not what is required, but this is expected to improve in the future. General support to detectors, commissioning ROSs. The majority of ROSs already commissioned with detectors: all of the calorimeters (LAR alone is more than 50% of all ROSs).

So far with LAr have achieved 50kHz readout rate, still short of the 75kHZ (to be investigated with M5, then M6 when SCT is installed). Tests with very high data request rates (out of the ROBin) have been successfully achieved.

WP2: HLT farm and networks

HLT farms

Fred

No new equipment purchases in the last six months, and no plans for immediate future. Now updating the specs in preparation for the next round of purchases (SFIs and HLT farms). The farms are generally running very well; demonstrated that local file servers can meet the requirements. Determined that a single file server per rack is sufficient to cope with all the traffic for all 30 machines in the rack.

Grid running (Jiri): no news

Performance of new AMD quad-core machines (Fred)

News on AMD quad core performance announced (2 weeks ago); not yet clear whether will buy from AMD (2GHz) or Intel (2.3GHz or 3GHz?). Detailed comparisons still remain to be done, until purchase next Spring or later.

Networks

Fred

Networks have been running quite well at P1. Some issues identified under abnormal usage/circumstances (such as when using HLT farms for EB; switches cannot cope with high data rate). New switches still being added to the system.

WP3: HLT software

L2 tracking (slides by Erkcan)

Fake rate (IDSCAN) reduced by a factor of approx. 3 (7% to 1.7%). In the latest version of IDSCAN the only-one-hit-per-layer rule has been removed, to allow hits from overlapping modules. Increased hit multiplicity helps slightly with the z0 resolution (especially at high pseudo-rapidity). Will try to get latest version of code into 13.0.30.2.

Timing measurements / tracking (slides by Dmitry)

Measurements with 13.0.30. Compared to 12.0.6, some significant increase, mostly from data preparation stage (especially TRT offline ByteStream converters).

ATLANTIS monitoring (slides by Jiri)

Fully integrated in the trigger monitoring for M4. In M5 TGCs will be included as well. Next, will add functionality to allow changing the type of monitored events mid-run, rather than just before the run starts (as it is at the moment).

Antonella

Antonella asked how to best integrate physics and trigger, for more coherent way of working and higher profile. Discussion ensued as to how to make the UK physics and trigger communities work together more efficiently. Trigger and Physics meetings in the UK? The UK Physics meeting in January is a good opportunity to make improvements in this area. Not advocating creating a new structure. John, Ricardo, Tracey already interfacing between physics groups and trigger project.

HLT steering

Simon G

New version of steering implemented about six months ago. Implemented improved monitoring, error reporting, and more robust algorithms. Better support for topological triggers; new RoI overlap removal. Persistency/transient separation evolving, to allow 13.0.30 data to be read with future releases.

10^31 menus implemented, about 200 different trigger chains included (cf 30 in rel 12). Work is ongoing for the 10^32 menus. Need to work on tightening selection thresholds, prescale factors, measuring background rates, also keeping an eye on the execution time.

Steering has been extensively and successfully tested over the summer (including in technical runs and cosmic runs).

Trigger configuration software

Paul Bell, Simon Head

Used in the last technical run, for the cosmic slice. Teething problems identified and ironed out. Will need to re-align the database evolution with the algorithm integration group.

Trigger Tool is now largely a done & delivered product. A demonstration of the TriggerTool, in preparation for M5 will be happening tomorrow.

Trigger software validation

Simon G on behalf of Ricardo (slides by Ricardo)

Ricardo has been mostly busy with HLT software validation in preparation for release 13.0.30. (Also has been covering L1 sw, as Olga left and her replacement is not yet active.) Organized a half-day Trigger Validation Workshop at CERN, on 26 September (21 participants); workshop included tutorials, aimed at trigger slice validation representatives and was very well received.

Ongoing work (started at the BNL workshop) on TriggerDecisionTools, together with Simon G over the summer.

Release Schedule

Simon G

Release 13 now; currently in final iterations of release 13.0.30.

Release 14 in February (aim for improved performance: memory usage, speed, initialization).

Region Selector

Mark Sutton has completed the new interface of the Region Selector. It remains to make necessary changes to the RoIDescriptor and to migrate the algorithms to the new interface, initially keeping the size/shape of RoIs unchanged, so as to validate the new version, then people to start using the new features.

Serializer (slides by Jiri)

This is a library to encode/decode the objects of L2Result and EFResult. It was used in the last Technical run and worked fine. Analysis if saved data is now ongoing.

Next OsC

HLT presentation will include the 6 main deliverables (3 hardware + 3 software), which the OsC wishes to tick off next Spring.

Hardware deliverables; ROBIN Performance - make reference to the ROBin paper?

Software deliverables; use timing and memory leak plots from Jiri and Dmitry.

Finances

Fred

Requisitions: very little spend so far this year (mostly related to ROBin work).

Finances are sound. No requests for funds at the meeting.

Will request to roll over about £50k-60k for next year. Extra money still needed for ROBin work.

Travel

Budget still underspent. LTA spending on the increase, but still under control.

Next meetings

Next phone meeting: Wed 5/12/07 at 11h00 (UK time).

Next face-to-face meeting: Wed 30/1/08 from 13h00 (t.b.c.): this will be organized by UCL, but meeting at RHUL.

-- PedroTeixeiraDias - 04 Nov 2007