AF meeting 21 September 2006
Andreas Pfeiffer, Andrea Dell'Acqua(ATLAS), Pere
Mato, John Harvey, Rene Brun, Fons Rademakers, Philippe
Charpentier, Gabriele Cosmo, Vincenzo Innocente
- Line management issues (John)
- Action items
- RFIO and Castor plugins status (Fons)
- AA Internal review postmortem
- LHCC comprehensive review of LCG
- Releases status
- Status of LCG_47 configuration.
- Experiments feedback
- Platform issues
- Topics for future AA meetings 
version 1 (28/09/06 @ 21:55)
Line management issues (John)
- The task to port the CMS background and radiation program to the new version of Fluka will be starting soon.
- RFIO and Castor plugins status (Fons)
Fons says that the separation of the two plugins is done and available
at the HEAD of CVS. Pere asked is CMS would require a release with it.
The only thing they need is to have a CVS tag. Fons agreed to tag it as
5.12.00(x), where (x) is the next one.
AA Internal internal review post-mortem
- General discussion took place in the AF about the preliminary
outcomes from the internal review. It was felt that this review has
been very useful for many reasons. The main one is that this has been
an occasion to ourselves to take stock of where we are and what still
needs to be done. It has also been very useful to inform the
experiments and other projects of the progress that have been made
during the last 18 months. One aspect that needs to be dealt
carefully, which may affect directly people working in the
projects, is that the reviewers should stay with their findings and
recommendations at the level of identifying the problems or express
their concerns but they should not try to solve or propose very
concrete solutions. We think that is up to the management of the
project to find the adequate solution.
- The projects will study carefully the final recommendations when they will be ready.
- PROOF evaluation and commitment from the experiments. We had a
discussion on how to exercise and evaluate the PROOF analysis
model within the experiments frameworks and event data models. It was
felt that has been a positive outcome from
the review but it is needed better communication. Effort need to be
found to see what will need to be changed/adapted in order to make use
of PROOF. For example an attempt should be made on trying to
run the Gaudi event loop and access to the event
data from a PROOF session. We will need to came back on this.
LHCC comprehensive review of LCG
- The presentations for the LHCC comprehensive review of LCG for
the Applications Area should be posted latest by Thursday. The
allocated time is 20 minutes, therefore a summary of the project status
should fit in something like 15 slides.
- Status of LCG_47
- Castor client libraries. This new configuration should
include a new version of Castor. Andreas reported that in the last C5
meeting at IT it was recommended to deploy anew version of Castor client library as soon as possible (e.g. before the experiments launch big productions or data challenges).
They would like to deploy version 2.1.1-1 that is binary
compatible with version 2.1.0-X . It contains a change in the protocol
and the URL syntax to make possible compatible changes in the future.
In addition, it fixes a bug that a client could get the data from
somebody else by re-using the same port. What remains a problem is
that old versions of experiment software could still be running the old
client library if the LD_LIBRARY_PATH is correctly set to the library
that is part of the configuration. Philippe said that by mistake this
is not the case for LHCb.
- It was agreed to wait maximum 24 hours to define the configuration in order to include the new Castor.
- Dcache. The configuration should include the new version of Dcache 126.96.36.199 that is backward compatible. With this version
it will be possible to use NIKHEF/SARA for running LHCb jobs. Andreas can do the installation of this new version.
- SEAL. It will go as is since the bugs in the component model,
mainly related to multi-threading, have not been fixed yet. The is an
action on Pere to propose a migration plan of the
component model to CORAL.
- Tomorrow will start the re-build of the projects.
As agreed the compiler gcc344 will not be included for this configuration.
- ALTAS. During the review it became clear that the communication
with the ROOT team with respect the usage of ROOT for I/O was not
perfect. Rene claims that ALTAS could gain factors in the
performance with some more discussions among the experts.
- CMS. Vincenzo reported that a number issues for ROOT are
still pending but workarounds exists for all of them (e.g.. ref,
sMatrix, multi-threading, etc.) . There is an action on CMS to collect
the pending issues and enter them into Savannah.
- LHCb. Nothing special, waiting for LCG_47
- ALICE. Nothing to report.
- Tag name for SLC4 with gcc 3.4.6. Andreas would prefer to
keep the standard full name (3 digits for compiler version) and add
sym-link with the reduced name (only the two most significant digits).
It was agreed to go back to the original idea of defining the tag with
two digits for the new builds and create the links for the existing
- MacOSx will be done when SEAL 1.9 will be out and for the moment for the powerPC architecture.
- We agreed that Pere on behalf of the AF should make a
request to IT to support the same
platforms as the AA for client libraries of the mass storage interfaces
and middleware services (e.g. Castor, gfal, etc.). In particular this
is needed for MacOSX and Windows.
Topics for future AA meetings 
Sep 27: ROOT topics (file I/O optimization, precompiled
headers, etc. )
- Oct 11: Trivial FC?, ProxyRef?