AF meeting 3rd March 2011
Clemencic, Benedikt Hegner, Fabrizio Furano, John
Harvey, Peter Hristov, Pere Mato, Rene Brun, Stephen
Gowdy, Marco Cattaneo, Gabriele Cosmo, Antonino Sergi,
Peter Kelemen, Andrea Valassi, Pete Elmer [remote], Sebastien
Line Management Issues (John)
Report from the LIM meeting (Benedikt)
News from projects
1 9/3/2011 @ 17:00)
Management Issues (John)
- 8 new LD positions has been advertised in the PH department. One
of them is for the SFT group to coordinate the development and
validation of physics codes used in the simulation.
- 7 summer students has been assigned to the SFT group (3 for ROOT,
2 for Simulation, 1 SPI, 1 CernVM)
the 5 to 10 most important requests/requirements for each of the
projects (Simulation, ROOT, Persistency, SPI, CernVM, Multi-core) (ALL
- Pere was invited to the ATLAS joint CMB/SPMB
meeting. Expecting a written input expressing the requests/requirements
in a 'customer language'. Similarly for CMS.
- The current plans will be made available for the
next AF once the production JIRA instance that we are setting will be
accessible from outside CERN and open to everybody in read mode.
- Propose a
presentation to the AF of the ongoing work in restructuring of the
XROOT client code and improvements for WAN transfers from ROOT (Pere)
- Dirk agreed for a presentation at the next AF.
an answer to the Savannah feedback document produced by ATLAS (Benedikt)
- Pending. Benedikt informed us that the IT
department has started a survey on
issue tracking systems (Savannah, JIRA, Track, ...) triggered by the
many people asking IT to setup a JIRA service. So far, this has been
done more or less informally, and it will be repeated later in a more
- As mentioned the current list of 'stories' for each project will
be made available before the next AF meeting.
- A written formulation (in customer language) of their
requirements will be welcome from the experiments.
Report from the LIM meeting
The last LIM meeting took place 2 weeks ago (see
agenda for details). There was no meeting this week due to other
obligations. Here is a summary with the current LIM issues:
- LLVM compiler. It was decided to replace the current LLVM
compiler with Clang, which is the "LLVM native" C/C++/Objective-C
- LCG_60a configuration. The configuration is now ready. LHCb
is currently building on top of it and they are satisfied so far. Asked
by Pere, Benedikt explained the main causes of the additional delay
incurred in the release of this configuration.
- There was a ROOT bug and the fix took a couple of addional
- A suspected bug detected by LHCb turned out to be in their
- A very clear list of platforms in required well in advance in
order to avoid last moment confusion.
- In general better synchronization between different teams is
- Marco C. requested a better documentation of the actual contents
of a release. Clicking in some of the pages redirects you to
nowhere. The links should should be revised and made the process
very simple to find the actual changes (release notes). ==> Action Benedikt.
In addition, the database with the external software versions is not
working properly. This needs fixing.
- Work is processing in the deployment of the JIRA instance.
News from Projects
- Persistency Framework
- Interest of NA62 to use CORAL. In particular, within an
application that links Geant4 and CORAL together, which will unveil
probably an inconsistency in the version of XercesC used by the two
projects. This implies that we need to distribute (in AFS and
CernVM-FS) a version of Geant4 that is compatible with the rest of the
AA software stack. So far Geant4 was not part of the AA configuration
release because each LHC experiment is using a different version. This
may change with the new requests.
- Building the PF projects with LLVM/clang. Axel is helping on
it. Andrea said that perhaps would be
better to move the exercise to SLC6 due to the standard libraries
required by Clang. In any case this is a test platform.
- 2nd Geant4 development release of 2011 has been released (it
was scheduled for end February). It includes the light anti-ion
developments to be tested by ALICE.
- Geant4 Technical Forum will be starting after this meeting.
- Vincenzo mentioned that concerning new R&D opportunities,
it was clear in the AIDA kick-off workshop that
multi-core/multi-threaded issues are not relevant for future
experiments at this moment. These detector studies projects value
flexibility more than computational performance.
- There are ongoing discussion in OpenLab about new compilers and
platforms (icc, gcc4.6, etc.). AMD has announced a new set of
- Data compression. Good performance in data compression (zip
library) is very important for the LHC experiments. The data
compression time (during data writing) is a large fraction of the CPU
time of a program such as the reconstruction. Also the achievable
compression factor enters directly in the cost of storage. Rene claimed
that the Intel ipp library does a much better job in compression and we
should be looking at it. Vincenzo claims that the standard zlib is
as good as the Intel ipp
when compiled with the proper flags. ==> Rene will follow
with Andrzej Nowak from OpenLab
- Whole-node job submission task force. Pete reported on the
current status. He is trying to get the experiments to perform their
tests, but the experiments are not really responding. ATLAS has done
something but ALICE, LHCb have never given any information.
- Tonight a development line based on LCG_60a will go in
production in the
nightlies. The release based on it will be done in a couple of weeks.
- Sebastien reported that there is a bug in gcc 4.3.2 and they
would like to switch gcc 4.3.x. and he asked for the plans. In
principle it should be fairly trivial and not noticeable (CMS is
actually using gcc
4.3.4). We decided to install the compiler and use it in one of the
nightly slots and switch to it later if everything is OK ==> Action Benedikt.
- Trying to get their software based on ROOT 5.28a.
- In a more general discussion they would like to see what are
the plans for releasing ROOT in several parts. In particular, a core
and I/O part separated from the analysis tools. They find that a single
package makes it very difficult for version migrations.
==> Action Fons to prepare a plan and present it at the next
AF. Marco C. added that it would be nice to have a more formal
definition of what is patch and what
- They have an issue with the icc compiler concerning licenses.
They have built all the LHCb software to
run benchmarks, but the question is where are they allowed to run them
with the current license. Can they distribute the binaries outside
CERN? They have asked the question to OpenLab and there is no
progress since one month.
- Testing ROOT version 5.28.00a. Several bug reports
are on the
pipeline [see clarification] .
- Peter reported the following news from IT:
- SLC 6.0 beta is now available.
- New issue tracking based on ServiceNow.
- bash shell is the default unix shell for any new account.
from Peter Hristov added after the meeting.
Concerning the fixes in Root v5-28-00-patches: the only one added is
the build of static Root libraries, which it is quite trivial. The
problem we saw with the friend trees was caused by a feature that had
to be taken into account on our side: ROOT accepts trees and friend
trees with different number of entries and it is the User
responsibility to check the matching between them. Our analysis now
works well with v5-28-00a.