Minutes of T2UK phone mtg / 06 Dec 2006
Present: John Baines, Paul Bell, Gordon Crone, Dmitry Emelyianov,
Simon George, Ricardo Goncalo, Barry Green, Richard Hugh-Jones, Julie
Kirk, Nikos Konstantinidis, Erkcan Ozcan, Pedro Teixeira-Dias,
Thorsten Wengler.
Apologies: Andrzej Misiejuk
Minutes of last face-to-face meeting
No comments.
News
Oversight Committee meeting (NK)
There was an ATLAS
OsC meeting last month. Nikos reportde the status
of the HLT UK project. The committee raised three main points:
1) request for clarification of our statement that, in view of latest
announced CPU developments, we are confident we can meet the
processing targets, given the return of the deferred funds. A
detailed reply was prepared, circulated for comments, and has now
been sent to the
OsC.
2) request for more precise specification of dates, milestones and
deliverables for the software part of the project. (To be addressed
at the next
OsC meeting.)
3) request for more specific project completion metrics. Namely, refer
to documents details of a spec, rather than just state that
"performance is according to specification". (To be addressed a the
next
OsC meeting.)
The next ATLAS
OsC meeting will be on the 14th May, 2007.
Large Scale Tests (FW)
Tests are ongoing, with emphasis on the database configuration and
access. Progress according to expectations. Currently the tests are
being run on a few hundred machines. This will be scaled eventually to
1000 machines.
Technical Run (GC)
There was a serious glitch due to a power outage last Monday --
caused a loss of the neutral phase going into one of the racks, when
they were brought back up after the power cut. This affected eight
machines and one switch, which are now out of action; the SFIs are OK.
Progress Reports
ROBin (FW, BG):
Now have an almost finalized list of ROBins for the next (final)
production: replace 16 faulty boards (7 from UK production + 9 from
German production) + 30 needed for detector testbeds and new
detectors. Order in preparation, with some urgency, as processors will
no longer be available next year. Manufacture will be at Gebauer, in
Germany.
BG: The issue of the development of a new ROBin (with PCI express
interface) was raised last week at CERN. RHUL is interested in
participating: similar work is already ongoing at RHUL, for the CALICE
DAQ.
BG gave details on the faulty UK ROBins: not clear whether problems
are due to faulty chip manufacture, or if problem has been induced by
static or mishandling of the boards. Slow/inefficient communication
with CEMGraft, as they are not replying to e-mails.
BG: the ROBin development team (RHUL, NIKHEF, Mannheim) met a few
weeks ago at CERN to set up the system to test cards remotely (as well
as locally).
ROS software (GC): porting old code to ERS (ongoing); expect new
internal release (tdaq-1.0.7) before the end of the year. The next
public release should be early in the new year (tdaq-1.0.8).
HLT farms and network
FW: a new test machine (dual core, 3GHz) has arrived in
CERN. Testing will probably be starting next week. Will try to get one
of the recently announced quad-core machines by Intel, for comparison.
RHJ reported on recent work on
-SFO evaluation tests, remotely, at CERN
-ELONEX machine perfromance
-Disk and RAID performance: multiple stream read/write performance is
still below ATLAS targets. However, tests by others do not show a
similar drop when using multiple streams. The difference is not yet
understood.
TW: Medium scale tests are ongoing. Regular reports at the pre-series
meeting; slow down of this effort because of LST effort. Paul Bell has
now switched from SCT to HLT and is participating in the Technical
Run. New HLT PDRA position at Manchester, expected to start in January
(interviews next week).
Trigger Configuration: L1 is essentially complete. Full functionality
is there for HLT now, but need to complete the list of implemented
menus. Well received in the review, valuable feedback given. Review
report to come out soon. Trigger Config Tool already being used in the
LSTs.
Core Software
SG: 12.0.4 - significant improvement in terms of additional
functionality wrt previous release.
L2 tracking
EO has worked for the release of the code for 12.0.4. Now going
over the test results from Monika and Dmitry regarding the efficiency
of IDScan (namely for misaligned data; frequency of missing space
points in e.g., pixel vs SCT, etc).
DE is looking at data with pile-up; first tests; some inefficiency
observed (3-4% drop) in the presence of pile-up (if no pile-up, no
change wrt previous results isobserved)
L2 track truth association: bug found recently; RG will be testing the
fixed version soon.
Finances
No bids. Financial report at the next face-to-face meeting.
Next meeting date
Need to change date of next face-to-face meeting (previously booked
for 24 Jan) due to a clash: It was agreed to have the meeting at UCL,
a week later, on the 31st January.
The following face-to-face meeting was scheduled for Tuesday 24 April,
2007.
(Minutes taken by PTD)
--
NikosKonstantinidis - 29 Jan 2007