Core Minutes 1/31/2006
Present: Joanne Bogart, Toby Burnett,
Johann Cohen-Tanugi,
Richard Dubois, Dan Flath, Riccardo Giannitrapani, Berrie Giebels,
Tom Glanzman, Navid Golpayegani,
Heather Kelly, Michael Kuss, David Landriu, Bryson Lee, Julie McEnery,
Igor Pavlin, Chuck Patterson, Leon Rochester, Tom Stephens, Tracy Usher
- DC2: Kick-off is
imminent
Sky model (Julie) It's on the verge of complete, just undergoing
final checkout: looking for duplicate sources, verifying that distribution
is ok, etc. There will be plenty of interesting things for people to find.
The whole process has been more arduous than expected, but should be finished
within hours.
|
I/O problems (Richard) When Interleave is on and there are large
numbers of jobs running, they become I/O bound. Clock time is running
a factor of 2 or 3 larger than cpu time, whereas the ratio has generally
been much closer to 1. The
nfs server gets overloaded, in spite of efforts
to move file-writing to local disks. (Julie) What about file reading? (Toby)
Code for reading during Interleave is uncomplicated and always closes
one file before opening another. (Richard) Furthermore, these files have
been moved to afs. (Julie) Might trying running jobs with just
Interleave. (Richard) We write last event seen to a log; perhaps this is what's
causing the problem. |
 |
- MRStudio (Riccardo) The overall
scheme—basic API—is working and he has used it to wrap CVS and
the more important parts of CMT. He's dealing with compile issues now.
Even if you're not interested in
more
details you'll like the illustration.
- FSW data coming: (Heather) Flight Software (FSW)
formatted data is expected by the weekend or early next week. We need
three new libraries:
- CCSDS Packet Retrieval library concerns the packet format. [In case you're
curious, as I was, CCSDS stands for Consultative Committee for Space Data
Systems. Somehow I don't feel enlightened. ed.]
- DFI (stands for Data Flow Interface) This is a parser which
allows us to identify and access the various kinds of information (physics
data, charge injection data, context information) in the data stream.
- Merging library Bryson's code to deal with duplicate packets,
time ordering of packets, etc.
The context information is new. It comes in different flavors, depending
whether the accompanying data is regular physics event data or output
of a calibration (and if the latter, it depends also on the subsystem).
There are also new bits concerning time. Eric C. will be helping Heather
in this effort.
- ISOC code migration: (Richard) had a meeting
with concerned parties and there is agreement that the ISOC code should
be moved to the Offline CVS repository and that we should make use of Release
Manager and other tools as appropriate to build, maintain and validate it.
(Bryson) He believes his Merging code is compatible with our CVS conventions.
Expects to treat CCSDS and DFI as externals, at least for now. Remainder
of ISOC code is currently closely coupled to Online. (Richard) volunteered
Heather to help with issues of package organization, etc.
- G4 upgrade: (Tracy) Geant4, v 8, requires
CLHEP 1.9.2 or later, which uses namespaces, which in turn means lots of
source code changes for us. There is a temporary work-around, but we
might as well get the unpleasantness overwith now. Tracy accordingly
modified 33 packages. He could then build, but the code wouldn't run
because there was a change to the form of the physics list. This, too,
has now been dealt with: G4Generator runs! Tracy is branching all
affected packages.
- Gaudi upgrade: (Heather) The newer Gaudi
version (v18r1—it must be 50% better
) also needs a modern CLHEP, so
Heather will piggyback on Tracy's work.
So far she has gotten the new Gaudi to build on Windows and
Johann has done the same for Linux.
- RM: (Navid) Apparently ROOT the gcc 3.2.3
binaries of ROOT can't be used with 3.4.3; we'll have to rebuild ROOT
with the newer compiler. In the meantime, lots of test programs in the
3.4.3. builds are failing.
In order to support GlastRelease on the Mac, we need binaries
for Gaudi and G4. Our current Gaudi version (v12r0) fails to compile on
the Mac (also with old gcc compiler), so Navid will hold off until
Heather has finished the upgrade.
Toby encountered a problem with RM and libraries on Windows. RM was not
copying static .lib files to the /lib directory, so that it was not
always possible to link against prebuilt releases. Since the meeting
Navid has fixed it.
- Pipeline: (Dan)
Archive stuff should be ready for testing by the end of the week; this
is the highest priority right now. Will move to Navid's batch interface
before long, after it's had some stress testing. (Richard) We're beginning
to use up our disks. He'll order more, but it will take a month. (Anders)
Reprocessing is done; it didn't take as much disk space as expected.
Double32 really helped.
- Data Server: (Igor) Speaking of disks,
u21 and u23, used by rsync, are full.
Branch filtering—facilitating selection of limited
output—is running on the test server,
glastlnx05, but it's ready for production.
(Richard)Data Server dictionary should be updated to include the CTB
variables; people will want to use them.
- Travel: Richard will be in San Diego
next week, attending an LHC review. Tom G. will be in Oregon.
J. Bogart, Last Modified:
01-Jun-2010 15:46:58 -0700
previous
