Core Minutes 8/22/2006
Present: Joanne Bogart, Toby Burnett,
David Chamont, Jim Chiang, Dan Flath,
Tom Glanzman, Navid Golpayegani, Heather Kelly, Michael Kuss,
Bryson Lee, Francesco Longo, Chuck Patterson,
Igor Pavlin, Tracy Usher
- Resource monitoring: (Navid)
He has been investigating an open source product,
Nagios,
used already by the SLAC Computing Center, which
looks like a winner. He believes we
won't find anything better. The one thing it can't do is to monitor
afs partitions, because of the need to get an afs token. We could use a
hibrid solution for this, as Tom suggests: a cron job can do the
real monitoring, as is done at present, and write a file of results.
Then nagios can look at the results.
Navid has nagios running now but output can't be seen outside the
SLAC firewall. For real use we will most likely put it on a new
webserver under our control.
(Richard) Question for Bryson: how should this be integrated with other
(LAT) monitoring? (Bryson) People concerned with this
sort of resource monitoring are a separate bunch from those watching LAT
performance.
- FASTCOPY glitches: (Bryson)
- The u23 disk filled up.
Fix: Direct copies
to a different disk, u37. FASTCOPY behaved properly: once the disk
was full, transfers stopped in 'pending' state; they resumed when
space became available.
- Poor FASTCOPY database performance on the Mobile Rack.
Fix: "Vacuum" the underlying
PostgreSQL database. These databases
needed occasional maintenance for good performance.
It hadn't been done for a while, so operations were extremely slow.
- Hand-off review (Richard) Generally speaking,
it's in good shape. Bill requests runs for the standard data sets with
new classification trees. (Toby) Let's wait for a proper tagged new
release; some packages need to be upgraded. One is
G4Propagator. It has some new exception handling
for 'stuck' particles which Tracy put in.
(Toby) There is still a problem on Windows with creating
large (> 2 GBytes) ROOT files. (Heather) It can be done with ROOT 5.10 with
a suitable TTree call.
- Beamtest (Michael)
- There was a problem with
recent calibrations not being found, causing jobs to fail, which has been
fixed. (Francesco) The calibrations had been registered in the (metadata)
database, but the files had not been copied to SLAC. Copying the files
fixed the problem.
- We need a geometry update for some runs.
- (Warren, Francesco) Some proton runs are failing in the Pipeline.
This might be
fixed by using the new version of
G4Propagator; it should be incorporated in
the next Beamtest release.
- Recon extrapolation makes an incorrect (for the Beamtest instrument)
assumption that trackers and cals match up. The orphan cal is not
properly accounted for, so the calculated energy sum can be too small.
- DataServer backend: (David) A set of tests has
been prepared in July. They are far from perfect, but good enough to check that
the code refactoring is not breaking the data server. The aim of the current
step is to join the peeler and the pruner code as much as possible. The top
levels are ok, and I am now going throw the low levels ROOT macros, especially
working on the centralization of information about the various kind of data we
would like to process (merit tuple, beamtest tuple, mc/digi/recon trees). After
each change I run the tests. Also, Tom is regurlarly running the head revision
together with web front-end and his own typical jobs. I think we will certainly
complete and install a new dataserver version this week, fully backward
compatible. Then we will start to make changes an discuss with web front-end
developers. [Thanks to David for providing this complete report after the
meeting. ed.]
- Visual Studio 8 (Toby)
In spite of misleading advertising on the part of the ROOT folks,
ROOT v5.12.00, as distributed, cannot be used to build and application using VCC 8.
One must instead use a custom-built version.
Initial performance measurements of a jet-finding algorithm
are not encouraging:
| |
cumulative elapsed times(s) |
| |
vcc7.1 |
vcc8.0 |
| Calc. stage |
Number |
debug |
release |
debug |
release |
| stable cones |
264 |
39 |
2 |
166 |
3 |
| midpoints |
472 |
132 |
6 |
557 |
10 |
| split/merge |
112 |
415 |
8 |
2523 |
18 |
J. Bogart, Last Modified:
01-Jun-2010 15:47:35 -0700
previous
