meeting 28th April
Antonino Sergi, Vincenzo
Cosmo, Marco Clemencic, Rene Brun,Pere
Mato, Lucia Silvestris, Liz Sexton-Kennedy
[REMOTE], Predrag Buncic, Pete
Elmer [REMOTE], Marco Cattaneo, Sebastian Binet [REMOTE], John Harvey,
Apologies from: David Rousseau, Andrea Valassi, Fons
Line Management Issues (John)
NeuroBayes license (Pere)
Report from the LIM meeting (Benedikt)
News from projects
version 3 (4/5/2011 @ 17:00)
IT has informed us that that Fabrizio Furano has kindly agreed to
replace Peter Kelemen as representative of the IT services, in addition
to his role as representative of IT-GT group. He starts in this new
role as of today's AF meeting.
Management Issues (John)
- A short-list of candidates for the LD simulation job has been
prepared. A selection board will be organized in a time scale of about
- One Technical Student has been granted to the SFT group to work
on the deployment and monitoring of virtual machines and PROOF in a
- An additional summer students has been allocated to the group.
- A request has been put forward to the Fellow Committee to fund a
Fellow from the EU AIDA project.
- Install the gcc 4.3.x
compiler and use it in
one of the nightly slots
- Unfortunately the problem is that the distcc
cluster does not know about it and it cannot be used in real
- No real progress on solving this distcc problem.
Fabrizio will follow-up this one and try to raise its priority in IT.
- SPI-33 Provide
an answer to the
Savannah feedback document produced by ATLAS
- New IT
contact person replacing Peter Kelemen (Pere)
- Date for
the next multi-core and virtualization
- Pere mentioned the new format and emphasis that Ian
Bird wants to give to the GDB meeting and the possible overlap with the
desired purpose of this workshop. It was felt that the workshop
dedicated to virtualization and multi-core is still needed to conclude
with the current R&D activities. Action Pending.
Pere asked John to summarize the long thread on the NeuroBayes license
issue and look at options for how to proceed. John started first with a
short history to make sure we all understand the context for the
discussion (looking at minutes of previous AF meetings)
The options available:
- April 1st 2010 - Topic brought to the AF by one experiment that
felt this was an issue affecting several experiments and clearly should
be discussed together. Conclusion: SFT was asked to conduct
negotiations with company on experiments behalf.
- April 15th: report at AF stating that verbal agreement reached on
conditions of initial use. Installation at CERN - nothing to pay,
institutes wanting to install training component would pay a minimal
- April 29th 2010: N was installed in AFS by SFT for evaluation
purposes. Experiments' prerogative to use in production if they wished.
Some AF members expressed reservations (Rene and Pete) : 'no objection
to evaluation of tool, but verbal agreement should not necessarily form
the basis of any final agreement'
- In the meantime, ATLAS as well as some CMS and LHCb groups have
- March 29th : Pere received a message from company stating that
more and more people were using the CERN installation. Since it was ~1
year since evaluation started they would like use of NeuroBayes at CERN
to be covered by an agreement. Difficult to manage agreements
separately for each user, would like to establish a general license
agreement covering all CERN computer users to reduce the paper work.
- We asked the CERN lawyers to look at the agreement to see if any
there would be any legal objections to signing it. They suggested some
changes which the company agreed to e.g. to protect against using the
name of CERN and logo when promoting the tool for commercial use.
- April 7th: this revised agreement was sent to the AF and the
experiments were also asked to comment on the contents of the
agreement. The positions of the experiments were presented in the
previous meeting and are summarized in the minutes.
- ATLAS/LHCb OK to sign
- CMS: not used in production but no objections to CERN signing
the agreement for benefit of other experiments
- ALICE: not using it, advises that CERN should not sign this
type of agreement
- Given the comments made in the recent email thread another
discussion needed before we can conclude on this matter.
If we decide on 2, only the experiments can decide what conditions are
acceptable. If no agreement on one of the above options:
- Sign version of the agreement agreed by the CERN lawyers
- Attempt to re-negotiate terms of the agreement that give more
favorable guarantees for long term use of the tool (e.g. availability
- Terminate use of NeuroBayes at CERN and remove the software
4. Request every group using NeuroBayes at CERN to sign
an agreement and we will remove it from the common AFS area
- Pete re-stated his position that, although CMS is not using the
package and does not plan to use it, CERN should not sign such
agreement because this undermines the position of CMS in case they want
to negotiate more favorable conditions (basically having access to the
- Marco re-stated that there is no specific agreement between
<phi-t> and the LHCb collaboration. NeuroBayes is part of the
software of LHCb and removing the software would cause a major
disruption. In addition, license agreements are signed by
'institutions' and not 'collaborations' therefore it needs to be signed
by CERN and is unacceptable that another experiment can dictate what
software LHCb has to use.
- There was general consensus that having access to the source code
in order to make recompilations and quick fixes would be an advantage
for everybody. In addition, it would also provide better long-term
guarantees in case the company disappear or changes dramatically the
use conditions. It was agreed to start a negotiation of a new clause
granting access to the source code under certain conditions. ==> Action Pere
Planning Preparation (Pere)
- 12 new 'stories' has been added after the first batch of 70
'agreed' last meeting. They will be reviewed at next AF meeting.
Report from the LIM meeting
- The LCGCVS migration didn't work properly. There has been a
problem about the load balancing and out-of-sync nodes, which caused
random checkout failures for the PF code. The remaining repositories
will migrate soon as well.
- Build cluster will get some regrouping soon. Currently test
machine running with new 'quattor' profile.
- Maybe a 60c will be requested, triggered by CORAL improvements.
Still to be discussed internally.
- The first iteration in gcc 4.5.2 shows many needed header
cleanups, as implicit includes have been removed. In addition, a new
gccxml version was needed.
- On project organization side, ATLAS plans to reduce the number
of projects from ~10 to ~4, to fix for too long runtime paths.
- A new CMT version might be due for early May. To be clarified.
- Needing a new expat in windows. A new frontier dependency in
GAUDI. No windows build exists. Is it supported at all?
- Concerning the future for the Windows platform. It was mentioned
the interest of the GAUDI users for it, similarly to the interest of
ROOT and Geant4 users.
- Benedikt reported on the recent problem of the RELAX installation
disappearing from AFS. It was recovered from backups.
News from Projects
- The ROOT team prepared an Easter development release:
- One of the visible features in this development release is a
new default style for the canvas graphics. The new style is called
Modern and is a synthesis of different custom styles used by several
large experiments. This new style is a work in progress, please leave
feedback on the forum with your comments on this style. It is the
default style to get maximum user exposure. The old style 'Classic' can
be selected explicitly.
- CernVM (Predrag)
- Released version 2.3.1 of CernVM to fix some Amazon
- Predrag reported briefly on the ongoing tests of the CERN
private cloud infrastructure (lxclould)
- Under development to include EOS (disk cache) as part of the
- Multi-core (Vincenzo)
- Vincenzo announced that Paolo Carlini (gcc standard libraries)
will be visiting CERN from May 12th to 15th. He will give a
seminar on the Friday and he will be available for discussions.
- Simulation (Gabriele)
- The new April development release of Geant4 has been made.
Among the features included, to note reviewed de-excitation hadronic
model and deployed operational rescattering in Bertini cascade.
- Marco reported on a long standing issue on memory usage by the
persistency layer (POOL+ROOT). This is now a show-stopper for some of
the LHCb data processing. They are validating the use of ROOT directly
skipping the POOL layer. CMS has already done a migration in this
direction some time ago. Rene is happy to help Markus.
- Withdrew the request for a LCG_60c, which would include some
CORAL improvements, to avoid the introduction of a binary
incompatibility between offline and online. They require compatibility
until the end of 2012.
- The validation of ROOT 5.28 is not yet accomplished.