The Architects Forum (AF) consists of the experiment architects and the
applications area manager (who chairs it), plus invited participants.
Architects represent the interests of their experiment and contribute
their expertise. Through this forum the experiments participate
directly in the planning, management, and architectural and technical
direction of applications area project activities.
The AF decides the difficult issues that cannot be resolved in open
forums such as the applications area meeting. Most of the time, the AF
will converge on a decision. If not, the applications area manager takes a
decision. Such decisions can be accepted or challenged by the experiments.
Challenged decisions go to the LCG Management Board (MB).
The AF meets every 2 weeks or so. Standard slot is Thursday 14:30-16:00
in 32-S-C22. Minutes of AF meetings are public and are posted
here. The AF uses the (closed) mailing list firstname.lastname@example.org
to communicate. Anyone can send mail to this list. The agendas
the AF meetings are also public.
Links to complete AF meeting minutes and summaries of the primary items.
- LIM report: use of Jenkins, releases 67a and 67b, Savannah to JIRA
migration, Coverity, Grid client software
- CernVM-FS strategic for LHC: Version 2.1.17 is undertaking the last
steps on beta testing.
- CernVM 3.0 released in January and very positive feedback from LHCb.
- New structure of the PH-SFT group and projects
- Discussed the HEP Software Collaboration initiative
- Release plans for the experiment software for 2014
- Roadmap of cvmfs deployment and end-of-life for cvmfs 2.0. The
conclusion is that sooner the better.
- JIRA service and Savannah migration status.
- JIRA service and Savannah migration status
- Reports from the projects (ROOT 6, Geant4 10) an experiments.
- SFT staff movements
- SPI: LCG_66, Neurobayes, infrastructure, savannah migration and
- News from the projects: status of ROOT6-beta.
- Announcement and preparation for the ROOT planning review. Scheduled
for June 14th.
- SCL5 gcc46 will be the only one maintained for SLC5. All the others
will be dropped (including 32 bit)
- Genser integrated within the 'externals'
- Quick summary of the ROOT user's workshop
- Discussion on the release plans for ROOT 6.0. The experiments would
like migration instructions as soon as possible.
- Discussion on support platforms and the strategy for reducing the
- Information on the ROOT user's workshop.
- Summary of Annual Concurrency Meeting held at Fermilab.
- Fixing the contents of LCG configuration 64d. Full support for SLC6.
- Status of ROOT 6. De-scoped the functionality and delayed the
- It was agreed to move the LXPLUS alias to SLC6 at end of April.
- Reviewed the long-term planning of platforms and versions. See the
details in the minutes.
- MC Generators will be integrated with the set of LCG externals using
the same build and deployment methods.
- Status of ROOT 6 release. First candidate release expected middle of
- Status of ROOT 6 release. First candidate release end of January.
Final release few weeks later.
- Migration of Savannah to JIRA. Scripts are ready. Experiments will
be contacted to plan the migration.
- Status of the MAC build service provided by IT
- Finished building all the externals with gcc 4.7.2 (compatible with
- Production platform for next year is SLC6 woth gcc 4.7. Experiments
still using gcc 4.6 for the transition.
- ROOT 6 release is being delayed by some weeks. For information at
the next meeting..
- Status of the Geant4 and ROOT major releases.
- Announcement of the ROOT workshop on 11-14 March 2012 in Saas
- Progress in the experiments to validate new versions and platforms
(SLC6, MacOSX, gcc4.7, etc.)
- Virtual meeting with short reports from experiments and projects.
- Feedback from LHC experiments on the ROOT 6 plans.
- Reported incident on ROOT server.
- LCG_63 configuration is in preparation; it includes ROOT version
5.34 and SLC6 as new platform.
- Reported the status of the Mac build service request to IT.
- Current release plan of ROOT was presented. Feedback requested for
- Request for a MacOSX build Service in IT has been submitted.
- Preparing configuration LCG 63 that will include binaries for SLC6.
- New JIRA service is now available. The idea now is to close the
creation of new projects in Savannah and re-direct people to the new
service. Feasibility study of the migration of Savannah to JIRA
- ROOT and Geant4 finalizing June releases.
- Presentation on the latest progress in the XROOT client library. New
notes taken during the discussion.
- Concerning the adaption of gcc 4.7, it was decided to start
experimenting with c++11 option first, and do a single build of all
packages with gcc 4.7 with c++11 turned on.
- Plans for ROOT releases 5.34 (in May) and 6.00 (in November) were
- Programme of work discussion. The main JIRA stories for each project
associated to the upcoming releases has been presented and discussed.
- Configuration LCG_62b in preparation. I will include new
version of Neurobayes and ROOT 5.32.2. New patch for LCG_61 will be
also required by ATLAS.
- Discussion on the needs by the experiments of 32 bit binaries for
recent LCG configurations and agreed strategy for providing them.
- Helge Meinhard summarized the status of the migration of the
interface machines of CernVM-FS from the pilot setup in PH to IT.
- Status and plans for the new JIRA service in IT.
- Reported on the new series of meetings of the Forum
on Concurrent Programming Models and Frameworks.
- Discussion on the priorities for the supported platforms (os and
compiler versions) for this year.
- Started to introduce the plans for ROOT, Geant4, CernVM. Details
will be needed to be added to JIRA.
- Summary of the ROOT I/O workshop (slides)
- Windows platform is now declared a non-production platform
- gcc 4.6 will be the next version of the gcc compiler supported the
the AA packages.
- Release candidates for ROOT and Geant4 are made available to the
experiments for testing.
- Status of configurations LCG 60d and LCG 61a.
- Notable progress of the CVMFS deployment in IT.
- Discussion of the deployment of gcc 4.6 and the interest of it by
- Presented and discussed initial ideas for a proposal of development
of a new data processing framework to fully support concurrency (slides).
- Experiments reported on their needs for the Intel icc compiler.
- Reviewed the stories for the persistency framework project (PF).
They are available
- Fabrizio updated us on the transition towards EMI and its
implications for the AA.
- Released LCG_61 configuration including ROOT version 5.30, updated
some Python packages, and some cleanup on a number of packages.
- Reviewed the list of 'stories' scheduled for the June releases
- Scheduling end-of-year releases. A late November release of ROOT
with candidate releases from the middle on November would
probably work for ATLAS and CMS.
- New CernVM release is due for today. This is using the new build
system (not using rPATH anymore).
- We will discuss the PF 'stories' at the next AF meeting.
- The dates for the next AF meetings until the end of the year
are July 7, August 18, September 15, October 13, November 10, December
- Agreed to build ROOT 5.26.00f with the xroot fix to avoid high load
on the Kerberos (KDC) service. ATLAS can extract the couple
of binaries for their ad-hoc deployment on critical cases.
- Decided to choose gcc version 4.5 as compiler for
the SL6 platform for the time being despite the added
complications for not choosing the native compiler version.
- Benedikt together with Nils Høimyr( IT/PES) have prepared a proposal
for an issue tracking service based on JIRA.
- Agreed that the AF meeting will take place once a month on a fix
time and weekday slot (Thursdays at 14:30) and a constant phase within
- The work for the completion of the ROOT schema evolution has been
broken down into a number of tasks, and their required
effort has been estimated. Fons will send the list of tasks to
the AF mailing.
- CwrnVM: demonstrator of executing Panda jobs on CernVM and having
the output to EOS.
- NeuroBayes. Only one experiment (LHCb) is actually using the
software and they ready to sign the license agreement. The text of the
agreement has to be adapted to reflect this. Agreed to
close this chapter at this moment.
- The migration from LCGCVS to CVS still not fully satisfactory. CMS
suffered a lot from performance problems, although it is becoming
- New ROOT patch 5.28.00d was released in urgency during the weekend
under the request of ALICE.
- CMS requested to have an evaluation of the effort needed to make the
ROOT schema evolution mechanism usable.
- Fabrizio Furano has kindly agreed to replace Peter Kelemen as
representative of the IT services.
- John summarized the long thread on the NeuroBayes license issue and
look at options for how to proceed. Agreed to start a negotiation for
a new clause granting access to the source code under certain
- Both ROOT and Geant4 teams have prepared new development releases of
- LHCb reported on a long standing issue on memory usage by the
persistency layer (POOL+ROOT) being a show-stopper for some of the
LHCb data processing. The pursued solution would be to skip the POOL
layer and go directly to ROOT.
- The migration for the CVS repositories to the AFS based service for
the projects CMSSW, COOL, LCGTEST, PF is scheduled for Monday 18th of
April at 9:30.
- NeuroBayes software. Experiments provided feedback to the proposed
license agreement between CERN and Physics Information Technologies
GmbH. The conclusion was proceed with the signature.
- Agreed to mark the initial planning stories as 'AF-agreed' and to be
monitored from now on. Newly created ones will be added for discussion
in subsequent meetings.
- CernVM-FS was discussed at several meetings last week (GDB, ATLAS
software week, etc.). Overall very positive outlook and very likely
that the CernVM-FS will be deployed to all Grid sites.
- Discussed the response to the ATLAS Savannah feedback.
- Agreed with the deadline for comments or feedback from the
experiments to the active "list of stories" till next AF meeting
- Discussed AF problems perceived by the developers in the experiments
- Report on the xroot new organization and developments plans by Dirk
Duellman. He also informed us about the EOS
(xroot disk pool) demonstrator status.
- Agreed on a proposal from Fons on how to release packages such as
TVMA, RooStat and RooFit independently from ROOT.
- Requested the experiments to have a look at the list of 'stories' of
the current program of work plans and provide feedback by next AF.
- Initiated a discussion on what is to be considered a 'patch release'
in forthcoming LCG releases.
- Started the migration of the CernVM-FS repositories with the new
- Production service hosted in the CERN computer center not yet
- The current list of 'stories' for each project will be made
available before the next AF meeting.
- Configuration LCG_60a has been released and is currently in use by
LHCb. The main causes of the additional delays incurred in the release
of this configuration were analyzed.
- Second Geant4 development release of 2011 has been released.
- Initiated discussion on data compression algorithms and whether is
required to study alternative implementations such as the ones
provided by the Intel IPP library.
- Agree to install the latest gcc 4.3.x compiler and use it in one of
the LCG nightly slots.
- Reports from ROOT, CernVM, CMS and LHCb.
- Selection of the JIRA tool to support project planning. Pursuing a
production installation with definitive license.
- A number of actions has been identified in order to mitigate the
causes of the delay (about 2 weeks) in producing the last two complete
- Agreed to continue supporting Windows for the rest of the year (on
best effort) and let LHCb proof during this time that there is a clear
- Call to the experiments to provide the 5 to 10 most important
requests/requirements for each of the projects (Simulation, ROOT, SPI,
PF). This request for input has also been extended to include the
R&D projects (CernVM, Multi-core)
- CernVM version 2.2 was released before Christmas. There has been a
lot of feedback during the holidays.
- Experiments stated their plans for releases to the upcoming
- Geant4 has migrated to SVN successfully.
- Presentation and discussion on the priorities and Organization of
the PH-SFT group for 2011-2012.
- Experiments reporting on their plans for next releases of their
- Feedback collected by ATLAS in the Savannah service. The SPI project
will digest the information and come with proposals on what is
feasible to be implemented in terms of improvements.
- Advancement of end-of-year releases for ROOT and Geant4 from 2011
onwards. From the viewpoint of the experiments middle of November
would be a good compromise.
- Possible restrictions of the CernVM-FS service during the end-of-the
year CERN closure. The experiments will not be able to make new
releases of the software during these two weeks. General agreement
that all the experiments would be fine with this limitation.
- G4 started with the procedure of the final tests (grid jobs) to be
ready for the December release. Next week on Tuesday there will be a
special Technical Forum meeting for the LHC experiments
- New version of RedHat 6 is now available.
- CernVM-FS very positive feedback from Babar usage for conditions
- Geant4 Development release (ref 09) will be done this week.
This is the last one before the public release in December.
- Reported on the main points from the latest ITSR meeting.
- New bug fix release ROOT 5.26.00e has been tagged by request of
- Installation of the latest versions for Doxygen, swig has been
- ROOT production version 5.28 will be released on December 15th. It
will include the improvements in the I/O subsystem.
- Discussion on the need for the RELAX libraries for STL containers.
- Benedikt will be joining the AF from now on. He will be replacing
Stefan at the end of the year.
- Reported on the main points from the latest ITSR meeting.
- Released configuration LCG_59a. ATLAS satisfied.
- Requirements from the experiments for the new production version of
- Coverity training week is taking place this week. Good feedback in
general so far.
- Summarized the issue of the clash between the different
implementations of the gssapi library.
- Announced changes in the Geant4 Collaboration
- Status of the preparation for releases LCG_59 and LCG58d. The time
scale for these releases is few days.
- Chris Jones reports that Neurobayes is being used in LHCb. LHCb
would be interested to provide feedback on their findings which should
help on the currently ongoing evaluation process.
- ATLAS is integrating a release with LCG_59 (ROOT 5.26.00c) and Alice
is currently testing new 5.27.04 ROOT release which is planned to be
used on the analysis facility.
- Multi-core and Virtualization WS feedback. A document summarizing
the workshop is being prepared. It will be circulated to the AF for
- Plans for the Persistency Framework releases.
- The Coverity training session has been fixed. The engineer will be
at CERN from August 16 to 20.
- Preliminary information about the plans for next major releases from
the experiments, in particular their plans to migrate to new versions
- Discussed the possibility to migrate to Python 2.6. Windows of
opportunity and plans for porting the Python bindings of the LCG
- Virtualization and Multi-core WS organization: Confirmed
speakers from ATLAS and CMS. The agenda
is now completed.
- Announcement that the report
the SFT group review is now available.
- Release LCG 56g has been done on May 11. The main motivation has
been to fix the inconsistency of the expat package.
- The long standing problem is now completely fixed in the Oracle 11g
client (OCI, OCCI, 32 and 64 bit) .
- Proposed to enlarge the scope of the AF to include other
projects/experiments in PH needing support for software and computing.
- Neurobayes. Decided on the installation of this package in AFS from
where any lxplus node we will be able to use it for evaluation
- Release of version 2 of CernVM based on on Scientific Linux 5.
Collecting feedback before the production version is prepared.
- Presented an alternative packaging of all LCG software merging all
packages in a common 'flat' distribution tree with a very simple
setup. Interesting for communities such as Theory developers.
- Configuration LCG_56f as been released, including fixes
in POOL and COOL requested by ATLAS. Needed a new configuration based
on LCG_58 to fix problem reported by LHCb with LFC in CORAL, when
accessing the Oracle DB and overloading the server.
- Updated information on Coverity and Neurobayes.
- CernVM 2.0-beta is almost ready and should likely happen in the next
- Configuration LCG_56f was requested by ATLAS, it was build
yesterday and is ready to be used. It contains mainly some fixes to
POOL and new versions of CORAL and COOL.
- Deploying the complete AA software stack for Theory. An initial
version exists and feedback is positive. More details for next
- Preparation of the 2nd
on adapting applications and computing services to multi-core and
virtualization has stated. The dates has been converged to June
20 and 21 (1.5 days).
- Interest from the experiments to use the Neurobayes
software. Agreed that the negotiations of the conditions with
the company is done in common for all experiments under the auspices
of the PH-SFT group.
- LCG_58b was released for LHCb. It contains a change in the Reflex
interface package and some minor updates in the external packages.
- Work progressing on building a CernVM version based on SL5.
- Decided to hold a Workshop similar to last year one to review the
progress made into adapting the computing services to multi-core and
- SFT program of work residential meeting.
- The migration of the Savannah service to the new hardware was very
- Tests of the ROOT I/O performance for LHCb based on Rene's
- Released two new configurations LCG 58a and LCG 56e that
incluces fixes to ROOT plugin for dCache and reverting to Oracke 11g
- Agreed to incorporate test suites to test ROOT plugins such as
xrootd and dCache to avoid the recent problems encountered by LHCb
very late in their software deployment.
- Peter Kelemen reported about the upcoming RHEL 6 release.
- Report on the Oracle-Opteron client library problem unveiled by
ATLAS. A fix should be available by Oracle.
- Testing of ROOT 2.26 proceeds in ATLAS and CMS. The new LHC schedule
after Chamonix may have an influence on its deployment.
- CernVM development version 1.5 is being finalized, it includes the
new version of CVMFS. There is some interest in using the file
system in other contexts (distribution of calibration files, software
installations in T1, etc.)
- The LHCC computing review will take place Tuesday 16th February. A
status report from the Application Area has been requested and Andrea
will be presenting.
- Interest by all experiments to use the coverity tool. Agreement on
the proposed training that will give us unlimited access to the tool.
- LCG_58 configuration has been released.
- Experiments reported on their plans for versions/configurations they
are going to use for the next LHC running period. Version stability is
a common pattern.
- Better understanding of the SELinux problems with icc compiler and
- LHCb requested a new configuration based on ROOT 5.26. Agreed to do
it in a time scale of a week.
- Started to have pre-view with Python 2.6. Agreed to start
an evaluation of what will be required for the Grid middleware.
- CernVM: exploring the option of a commercial content delivery
network for the experiment software delivery.
- Re-starting the meetings of the multi-core R&D project.
- Had a look at the new organization of the IT department and
discussed the main changes in the groups nd how this affects the AF
and the experiments.
- Status of configuration LCG_56d that was recently released.
- Latest news on the end of the year releases from Geant4 and ROOT.
- CMS reported progress on testing latest version of ROOT.
- Clarified the reasons why the version of SVN client available in
SLC5 is different than SLC4.
- AIDA EU-FP7 proposal was summarized. Software work package (WP2)
proposes to develop a generic HEP geometry toolkit, and detector
independent reconstruction tools.
- ROOT version 5.26 to be released on December 15th. It will be nearly
identical in functionality to the release 5.25/04.
- Geant4 production version will be released on December 18th.
- Configuration LCG_57 was released in Oct 19th. It contains new
versions of all projects and is based on ROOT version 5.24.00b.
- Defining Configuration LCG_56d. Some externals will need to be
updated, and minor versions of the AA packages still waiting for fixes
from CORAL, POOL, and COOL. The initial urgency has been relaxed.
- Summary the status of the POOL project. The conclusion is that
POOL should stay in maintenance mode with the best strategy would be
to get the ROOT part out of POOL and move it to the experiment's
- Version 1.4 of CernVM was released end of October.
- Discussion on how the change the order in ROOT to take into
account the LD_LIBRARY_PATH first.
- The development version of ROOT 5.25.2 has been just released. Rene
summarized the main highlights.
- The new Oracle client library version 11.2 recently installed does
not solve the problem with SELinux.
- Discussion on how patches are applied on top of ROOT. Analysis of
the list of bug-fixes introduced in Root from 5.22/00a to 5.22/00d to
understand ways of improving adaption of new versions of ROOT by
- Additional SLC5 issues. The lxplus alias business seems now
converged. No new date has been fixed to when the switch to SLC5 will
- Started a discussion on how to release/deploy physics analysis
software packages layered on top of ROOT such as RooStat or RooFit. A
couple of actions identified. There was no conclusion.
- No need to create a new configuration LCG_56d with only
configuration fixes as it was originally requested by ATLAS.
- Decided to install and migrate to the recently released new version
of the Oracle thin client 11.2, which includes all the required fixes
and it also solves the problem with SELinux. The light version will be
- ROOT includes now a new optimization schema for the size of the
buffers associated to each branch in a tree.
- Confirmation that the additional bit to be enabled in SELinux
(allow_execheap) will be required as long ATLAS needs to run old
software and the problem with the Oracle client library is not fixed.
No clear time scale yet.
- The release of the LCG_56c configuration was done successfully and
well in time for the ATLAS releases.
- Agreed to prepare a new configuration LCG_56d consisting of only
LCGCMT modifications but no other package changes.
- There have been some feedback about the SL4 compatibility
meta-package. If anybody knows of something missing should inform
- The proposal to migrate the Oracle client version to 188.8.131.52
presented by Andrea Valassi was agreed.
- Inconclusive tests on what SELinux bits are required for Oracle
library to work on SL5
- Details of the contents for the new configuration (LCG_56c) have
been discussed and agreed.
- Discussion on the SLC5 deployment at CERN and follow-up of the
proposal of disabling SELinux.
- Additional SLC5 issues and feedback from latest GDB.
- Proposal for a LHC-wide tool to analyze the content of our
applications heap. There was general agreement that the kind of
information collected by the different tools is the similar. Agreed to
setup a group of experts and discuss the details.
- The agenda for the Workshop
adapting applications and computing services to multi-core and
virtualization is now complete.
- Update of the ATLAS status and plans for the migration from the CVS
code repository to the SVN one.
- Discussed and agreed the contents for the proposed new configuration
- Reviewed additional pending issues for the SLC5 migration and how to
improve communication. Missing recent update on the concrete plans of
deployment of it at CERN.
- ATLAS reported a significant reduction of the virtual memory needs
for 64 bit binaries such that they can run 64 bit applications on the
- Discussion on the CVS to SVN migration of experiment's code
repositories. Changes in the priorities and plans from the experiments
due to the new LHC schedule.
- SLC5 migration issues:
- CERNLIB libraries has been re-compiled and tested by ATLAS.
- New failures of the FIO provided SLC55 lxbuild nodes for the
nightly build service(s). Slow responsiveness.
- Relesed new version of gLite for SLC5 (64 bit)
- A lot of interest and nice feedback for both multi-core and
virtualization R&D activities at CHEP.
- Starting to prepare the workshop on adapting applications and
computing services to multi-core and virtualization.
- A new Geant4 patch for 9.2 was released last Monday. CMS was waiting
for it. Both ATLAS and CMS are now aligned to the same version of
- SLC5 migration issues:
- One of the biggest problem for the SLC5 migration is being the
slow acquisition and correct configuration of the lxbuild nodes.
- The CERNLIB libraries need to be re-compiled with -fPIC to avoid
the problem with SElinux detected by ATLAS.
- Experiment SLC5 migration status: CMS is prioritizing on the 64 bit
version first. LHCb is running SLC5 native code on the nightlies.
ATLAS is also currently integrating on the nightlies.
- The ROOT development release 5.23/02 was announced. ALICE is already
running on it in production. LHCb and ATLAS are currently testing
directly the head revision. A new ROOT web page based on the Drupal
content management system is in place.
- Additional SLC5 issues such as the cron/cvs patch, the Oracle client
fix, CERNLIB problem were discussed.
- ATLAS decided to upgrade the simulation to Geant4 9.2. The
production releases for ATLAS and CMS have been delayed for 2 and 4
- Discussed the request to include the KSM module as part of SLC5. No
conclusion was reached. The idea of having a Workshop to discuss
multi-core and virtualization issues was very well supported.
- Reviewed the list of pending issues related to the migration to
SLC5. The migration of the majority of the CERN resources to
SLC5 at the time scale of April is probably fine. The lxplus alias
could be changed sometime between April and September (before the
- Every experiment is interested on a transition to SLC5/64bit of the
Grid resources as soon and as short as possible.
- LCG_56 configuration was released last Friday. ATLAS has
already taken it and there are no problems so far.
- LCG_56 configuration was basically ready for release. Stefan
summarized the latest steps to be taken.
- The release of Boost 1.38 was confirmed to be ready by Monday,
therefore in time for be incorporated into the LCG_56 release.
- ATLAS reported problems with their candidate release. They have
appear after the bunch of fixes to ROOT has been made available,
together with a change in the API for CollectionProxy,
and that more experts are making final validations.
- CMS reported on their tests of the new ROOT and POOL. Things look
fine in the simple tests being done and all problems has been fixed as
far as they can see.
- LHCb is testing a lot of things at once (including new version of
Gaudi). They can run applications with the preview release of LCG_56.
- Concerning the deployment of SLC5, CMS has a problem with changing
the lxplus alias as it has been initially announced.
- Concerning the LLVM proposal, we concluded that we need to do the
validation of CINT7 in the next few months and take a decision on the
new project a bit later.
- Status of LCG_56 configuration. Reported about the reminding issues
needed to be solved before the release. This one is still expected by
the beginning of February.
- Agreed on a presentation for AA in the forthcoming LHCC mini-review
of the LCG. Pere will be the presenter and he will collect a few
slides from each of the projects.
- Dirk reported about the review of Castor that has been taken place
and the plans for its future. Notes taken during his explanations. For
more detailed information pointers to presentations about the subject
in various meetings and conferences has been provided.
- Status of releases: LCG_55c is done. LCG_56 (pre-release) is
ready. A number of issues have to be fixed before the production
release aimed at end of January, beginning of February.
- SLC5 migration:
- AF supported the proposal to not include CERNLIB part of SLC5
- A number of issues need to be understood about the differences
between RedHat provided gcc compilers and the ones currently in use.
- Understanding the deployment plan for SLC5. Clarifications will be
asked to IT/FIO.
- Agreed that all the LCG-AA software should be supporting the 2.X
version of CLHEP and not only the version 1.X.
- Reported the feedback from experiments on the LCG_56 preview (new
ROOT version). ATLAS a long way to go to provide feedback, LHCb well
advanced and good so far, CMS could not test latest previews, and
Alice fully validated.
- Reported on the status of the Grid client libraries for SLC5,
gcc-4.3 and Python 2.5 and the tests from the experiments.
- Decided to include new version of CLHEP (184.108.40.206) in LCG_56.
- Stefan reported that the migration of the ATLAS HypeNews items to
SharePoint is completed.
- Report of the progress of validating SLC5 by the experiments.
- A number of small problems are being studied.
- A new requirement was identified: to be able to build
SLC4-compatible binaries using SLC5 systems.
- The ROOT problem in the area of Reflex that renders unusable the
recently released development version of ROOT 5.21/06 was discussed.
- In the context of the multi-core R&D project, Vincenzo reported
on the encouraging results of the tests of "ksm" (shared memory kernel
- A new patch version of CernVM v1 is ready for release.
- Configuration LCG_55b (de-sealed) was finalized and announced last
- Finalized the contents and supported platforms for configuration
LCG_56. A pre-release should be made available towards the end of
November and the production quality one mid-January.
- Reported the progress for the introduction of SLC5. The ATLAS
compatibility tests on hold due to an AFS problem. The SLC5 platform
should be available in the nightlies from now.
- Configuration LCG_55b (de-sealed) is almost ready with some problems
in building COOL and other packages on VC7. The idea is to
include a new ROOT bug-fix version (5.18.00f) and have it ready early
next week. This is compatible with the ATLAS and LHCb needs.
- SLC5 software migration strategy
- Encounreted problems with SLC4/gcc-3.4 binaries in compatibility
mode (phase 1) with the SElinux module and possible memory leak.
- Agreed to choose gcc-4.3.2 as the official compiler for SLC5
(phase 3), together with python-2.5, therefore requiring support for
- Reported progress on Windows VC++ 9.0 and gcc-4.3.2 ports of
the LCG-AA software.
- The first release of the CernVM software, 1.0, is now ready and
should be available starting from October 31st, after 6 months of
- LCG_54h will be released this week. The de-sealed version (LCG_55b)
is not urgently needed and will be done later.
- The 2009 "production" version of AA software stack is requested by
mid-January 2009. This request should not be a big problem as long as
the integration tests by the experiments start as soon as a stable
version of ROOT is made available (end of November). The final
production version of ROOT 5.22 is foreseen by December 15th.
- Discussed the inclusion of SLC5 as new platform. A plan of action
with 3-4 steps have been defined. The major decision still pending is
to know if SLC5 binaries should be produced with the native compiler
gcc 4.1 or to jump directly to gcc 4.3.
- The two proposals circulated by Stefan (CMT tag name unification,
and procedure to manage software installations) has been agreed.
- Adapting to the new LHC schedule.
- Experiments would like to have the full software stack ready by
end of January, such that they can start building their
software that will be used in production during 2009.
- To avoid long integration times with a new version of ROOT,
CMS will build a release with the current ROOT development
- Geant4 9.2 release is scheduled for December.
- New configuration in preparation (LCG_55b). It will have some
some minor changes in the externals, include a separated release of
RooFit based on ROOT 5.18 and changes in the Grid packages.
- Conclusions from the SubVersion presentations in AA meeting.
Experiments will only plan the migration from CVS to SVN during the
2009-2010 winter shutdown.
- A new correction for Geant4 version 9.1 (patch 03) will be
released this week.
- Reported progress in the VC9 port. The main remaining issues are
been solved, but several errors are still observed at runtime.
- Small report from both R&D projects: virtualization and
- Discussed a new proposal for the distribution of the middleware
software. A pull mode, in which sites can decide what they want to
install and the possibility to have parallel installations with
different versions, was favored. Oliver will provide the tarballs to
SPI to see how these can be integrated with the rest of the AA
- The development version of ROOT 5.21/02 was released at the end of
August. Contrary to what was planned, this release does not include
the new automatic schema evolution.
- Some progress reported on the VC9 port. Problems with COOL and CORAL
has been found in a number of tests failing at runtime.
- A SVN session will be organized in the first AA meeting in
- Experiments have released the software versions that will be used
for the first data.
- CernVM version 0.7 was released. It comes with a new implementation
of the file system with some extra goodies such as the possibility to
mount the file system in read-write mode.
- Andrea Valassi replaces Dirk Duellmann as leader of the Persistency
- A presentation in a post-C5 meeting of a proposal for a new SVN
service and closure of the existing CVS service has caused some
reaction from the experiments. The post-C5 meeting was felt not to be
the correct method to communicate to the experiments. A presentation
of the proposal will be scheduled in a AA meeting.
- Configuration LCG_55a has been released. This new configuration
includes the new version for HepMC, which now can be the same for
ALTAS and LHCb
- All the experiments are releasing their software versions that will
be used for the startup. Only fixes to important bugs will be
incorporated. In addition, they are also preparing major versions to
be put together later in the year.
- Virtualization project. Released version 0.6 of CernVM
- Multi-core project. Progress in multi-threaded Geant4 and
investigating various memory allocators for multi-threaded
- The ROOT release scheduled for 26 June will not include the new
extension of the schema evolution. The introduction of the new schema
evolution functionality will be done by the end of August. The
production release is expected in December.
- Status of configurations:
- LCG_54g has been released, LCG_55 can go immediately undoing the
upgrade of the version of Boost that did not work straight
- Boost 1.35 will be added in the "dev" nightly slot to prepare for
its inclusion in future configurations.
- There will not be a new production version of Geant4 in June this
year. There will be instead a beta release (9.2-beta) which will
be made available for testing to users at the beginning of July.
- Preparing in parallel the LCG_54g configuration (including
ROOT 5.18/00d ) and configuration LCG_55 (including "de-sealed"
versions of POOL/CORAL/...)
- Reported of the ongoing effort for reducing the complexity of the
POOL package; summary of of series of meetings with representatives
from the experiments that has been held to know about their use of
- Decided for the moment to go with gcc 4.1 for the SLC5 platform,
which is the default compiler version.
- The new 0.5 release of CernVM is under preparation; the release
includes small fixes and extra packages on request from LHCb,
addressing SLC4/5 compatibility libraries.
- A new version of the specs for the enhanced ROOT schema evolution
has been made available. People should circulate it to experiment
experts through their distribution channels. There will be a
presentation to one AA meeting later but comments/feedback should be
sent to Rene directly.
- The software for the new configuration LCG_55 (desealed) is in good
shape. It was decided to make the release of it by the end of May.
Will need to maintain the two configurations 54 and 55 in parallel
until ATLAS will make the transition.
- New releases:
- Development release of ROOT 5.19/04 is now available. See release
notes for the details.
- New patch for Geant4 9.1 has been made available. See release
notes for the details.
- Some progress reported on new platforms SLC5, MacOSC 10.5 and VC9.0.
- The progress of Virtualization project (CernVM) was reported.
- Agreed to prepare a new LGC configuration fixing the problem of the
dCache installation together with a number of version changes in
external packages such as gfal, Castor, lcgutil, etc.
- Asked experiments if support for ROOT 4.x files by POOL is still
needed. This would allow some cleanup in RootStorageSvc, thus
simplifying its long-term maintenance. Waiting for ATLAS to give the
- LHCb and ATLAS would like to have one more complete release by June
with the 'de-sealed' version of POOL/COOL/etc, but no change in
the ROOT version. They do not plan to use the June release of ROOT in
production until next year, although tests of the improved schema
evolution will be performed in order to provide feedback.
- The kickoff workshop for the R&D projects was commented.
- New configuration with fixes to POOL has been produced (LCG_54d). It
includes also a new nersion of dCache.
- Large progress reported in removing the SEAL dependency (de-sealing)
of the POOL packages. POOL itself is completed. CORAL is not not
finished but very advanced. COOL still have very few things to be
- Changes in the Nightlies
- Building SEAL will be taken out of the "dev" slot.
- The "dev1" slot in the that is dedicated to patches to LCG_53 will
also be stopped.
- Building for SCL3 will also be stopped since no new requests are
expected from experiments.
- Requested by LHCb (Dirac) to provide a statically linked version of
Python in the release area.
- Circulated the very initial version of the note
describing the specs for the advanced ROOT schema evolution provided
- Initial discussion whether or not the experiments would migrate to
ROOT 5.20 (summer version) and what would be the work required for it.
A major release of the POOL projects removing the dependency to SEAL
is already scheduled.
- Compiler versions:
- Since the 'official' gcc 4.3 has been released, it was
decided to jump directly to it skipping 4.2. A preview in the
nightlies will be setup soon.
- For windows, despite of some concerns we agreed to move to VC 9.0
in a time scale of few months.
- Reported the status of the R&D projects. Finalizing the agenda
of the kickoff
- Reflex API change proposal. Need to have the documentation of the
actual API proposal to make things concrete and see what is the impact
of these changes would have in the client code. Header files for the
new API would be sufficient as specification.
- The actual development work for the combined RFIO client library
have not yet started.
- Two new bug fix configurations 54b and 54c had been released.
- All experiments have plans to converge with Geant4 version 9.1
in the long term. The June release of Geant4 may be only a bug fix
release on top of 9.1.
- The migration towards MacOSX 10.5 (leopard) has started with the
newly available build server. The platform tag was agreed to be
- The ROOT interface to the Qt GUI toolkit version 3 (Qt3) will not be
supported starting from next production release. Only Qt4 will be
- Agreed that Architects Forum will also monitor/drive the
developments for the two new R&D projects hosted in the PH-SFT
group. Both project leaders will be invited to the meetings and short
status reports will be given regularly.
- LCG_54a configuration was almost ready. It includes some changes
with respect LCG_54: ROOT 5.18.00a, CORAL 194, COOL 231, POOL
342, GSL 1.10. A new patch configuration will be prepared later to
include new version of COOL.
- Experiments are picking-up configuration LCG_54(x) for building
their production releases during the next weeks.
- Requested to experiments their support needs with respect the
version of the Qt GUI toolkit and its interoperability with
ROOT.Decided that if no reason is identified from ATLAS or CMS to keep
Qt3 support, then ROOT will move to Qt4 exclusively.
- Discussion about CORAL server development. Decided to go ahead with
the proposal and develop the prototype. This prototype needs to be
delivered quickly because of the urgent needs from ATLAS online.
- New configuration in preparation with latest bug fixes. ATLAS would
like to defer the release of LCG_54a for a week. This was agreed and
the target date for the
bug/fix release has been fixed for February 18th.
- John Apostolakis made a summary of his analysis of the effects the
DoE budget cut in SLAC has for Geant4. He also has written a memo with
more details and has been sent to ATLAS and CMS management.
- The release of LCG_54 configuration went well in general. It was
agreed that further and more extended testing will be needed in the
- There will be some changes in the nightlies with new slots for
patches of LCG_54 and new version of gcc compiler.
- Dirk presented the CORAL server proposal in last
AA meeting. The experiments should comment the proposal.
The decision eventually could be taken at the next AF.
- The production release of ROOT scheduled for next Wednesday 16th
assuming all pending problems are solved and no new ones are detected
by the ongoing experiment validations. A couple of days
later we should get the complete LCG_54 release.
- Python 2.5 has been validated and is going to be added into the
candidate configuration for release ("dev" slot in the nightlies)
- Agreed to introduce MacOSX 10.5 (leopard) as new platform after the
release of LCG_54
- Agreed to start working for gcc 4.2 after the release the LCG_54.
Eventually we will drop gcc 4.1 after complete software stack has been
validated with gcc 4.2.
- Suggested to have Geant4 and Genser MC generators as part of the
nightlies to facilitate the validation by the ATLAS.
- Witek Pokorski will be replaced by Alberto Ribon as project leader
of the MC Generator services from in February 2008.
- The release of the production version of ROOT will be delayed until
January 16th. Meanwhile a development release (5.17/8) will be made
available this week as the 'release candidate' for the experiments to
validate. Only bug fixes will be committed in the code repository.
- LHCb tested their software using the nightlies and found a blocking
problem concerning reading old data files. A fix has been prepared and
is being tested.
- Geant4 version 9.1 will be released on December 14th as
planned. The validation of the hadronic physics has been done
with 5000 jobs submitted in the Grid. It is about half the number of
jobs compared with last time mainly due to the reduced computing
resources with SL4 available for Geant4.
- A new request will be issued for the Castor (rfio) client libraries
for MacOSX and Windows. These libraries are needed when experiment
software is executed in these platforms for data analysis.
- The next LCG configuration (LCG_54) will be based on Python 2.5. It
was agreed to add Python 2.5 into the "dev" slot starting from
- Quick update on the ROOT status of schema evolution and support for
split mode. A specs document is still planned before the end of the
- Positive feedback from the last LCG Comprehensive review.
- The integration of the configuration LCG_54 (preview of the ROOT
production version) is progressing well. The "dev" slot in the
nightlies is slowly converging. The final list of changes in
external packages versions will be circulated for comments soon. New
versions of POOL, CORAL and COOL are foreseen.There will be an
announcement when the slot is completed such that the experiments can
do their validation of the pre-release.
- The conflicts between Globus distributed with the Application Area
and in the LCG UI detected by ATLAS seem to be solved with the
recent changes made in LCGCMT.
- The decision to upgrade to Python
version 2.5 was postponed until January when ATLAS would have
concluded their tests.
- Discussed how the 64 bit version
of the Grid Middleware in WN (i.e. clients) will affect the
- Finalizing the talks for the LCG Comprehensive review. The
presentations would need to be circulated to the AF mailing list as
soon as possible. Concerns expressed by the program of work presented
- Configuration LCG_53f has been made containing new versions of
CORAL, COOL and RELAX. This is the first time the release has
been build by the SPI team. There were same glitches that were solved
in less than 24 hours.
- Decided to make the nightly slot 'dev' fully working as a preview of
the December releases including ROOT production version 5.18. This is
to be used by experiments to validate the release. The support
for SLC3 and gcc32 will be removed from that release.
- Scheduled end of the year releases for ROOT and Geant4.
- Finalized agenda for the LCG comprehensive review scheduled for
Monday 19 November.
- Released ROOT version 5.17/04. Rene will present the new
developments in the next AA meeting.
- Configuration LCG_53d was finalized. Starting to work for LCG_54
that includes new version of ROOT.
- LHCb asked about the status for discontinuing the SEAL pluging
manager. The milestone was scheduled by end of November but this is
not realistic taking into account the available developers. The
current idea is to move the milestone to end of March 2008.
- Configuration LCG_53c was released with the new version of gfal
- Configuration LCG_53d will we released next week with a number of
agreed changes needed by ATLAS and CMS in POOL, CORAL and POOL. See
the minutes for the details.
- The proposal to support complex schema evolution in ROOT has been
circulated by the experiments to their experts and meetings to discuss
concrete use cases are going to be organized soon.
- Agreed on the agenda for the next AA session of the LCG
comprehensive review scheduled for Monday 19 November.
- Announcement that ROOT have migrated the code repository from
CVS to Subversion (SNV)
- The proposal to support complex schema evolution in ROOT was
presented and initially discussed. It was agreed that there will be a
round of discussions with each of the LHC experiments and POOL to
confront the proposal with the concrete use cases. The goal is to come
with complete specifications of the API by December 2007, and an
implementation by June 2008.
- Agreed with the next series of configurations:
- LCG_53c. Will be made as soon gfal 1.10.2 is made available.
- LCG_53d. A complete release of the MacOSX (intel) stack for all
projects, including requested bug fixes in COOL and CORAL by ATLAS
- LCG_54. Preview configuration including ROOT 5.17/02
- MacOS (intel) binaries are being build in the "dev1" slot on the
nightlies and they look in pretty good shape.
- The experiments will wait for ROOT development release 5.17
scheduled end of August to integrate it in their releases.
- Decided to prepare new configuration LCG_53b with the necessary
changes to cope with gcc 4.1 and MacOS(intel) and compatible with
current experiment's versions. Changes in versions of GCCXML, RELAX,
CLHEP are expected.
- Decided to discontinue MacOS(powerPC) and SLC3(debug) as binary
platforms for the next releases and introduce MacOS(intel) instead.
- One limited duration (LD) position has been announced in the IT/PSS
group to work on POOL/CORAL.
- Discussed the two R&D proposals that are being considered in the
PH department in the area of application software for the
- Started a campaign to make the nightlies for POOL and CORAL "green".
The MacOSX(intel) as a platform for the nightlies is coming soon.
- The configuration LCG_53
is almost ready. Only COOL is still missing, which is
going to be build very soon. ATLAS is waiting for it.
- Both ROOT 5.16 and Geant4 9.0 are about to be released. They will be
announce at the end of the week.
- Decided to prepare configuration LCG_53 including the new
version of COOL and we ready in one week.
- The installation of the Grid client libraries in the LCG external
area is now completed and has been tested successfully by LHCb. It
contains the SL4 and SL3 versions but the SL3 has not been certified.
Experiments should make sure that the entire set is deployed/installed
to the grid nodes and not do rely that the installed versions will be
compatible. No actions pending.
- There have been some troubles in the nightlies during last week, but
the system is stable again now. A slot with gcc 4.1 will be open.
- The production version of ROOT 5.16 due for June 27 will be
validated with only the LCG nightlies including Gaudi. AA projects
should check careful the results of the "dev" slot that is being build
nightly with the candidate release of ROOT.
- No major problems with the validation of Geant4 9.0. A candidate
release will be made available early next week to the experiments.
- The new release of COOL is expected to improve the performance
scalability for the IoV lookup. ATLAS would like to have this new
release of COOL at the latest the first week on July.
- A summary of the special meeting that took place among the
specialists on the integration problems of the DPM and
RFIO plugins for ROOT and GFAL was presented. The AF endorses the
conclusions and proposed actions.
- The integration problems of the DPM and RFIO plugins for ROOT
and GFAL were discussed. A document describing the problem and
possible solutions is being prepared by Flavia Donno. A meeting with
the experts will be organized to discuss the proposed solutions.
- A LCGCMT configuration containing interface packages to the new
installed Grid s/w in /afs/cern.ch/sw/lcg/external/Grid/ has been
prepared and is currently tested by LHCb.
- The list of POOL/CORAL tasks and their status was circulated by
Dirk. Having finished the ATLAS request, the development team has
started handling the CMS items.
- ROOT 5.14.00f patch version was released Wednesday morning. See
of changes. The ROOT pro release 5.16 is scheduled for June
- Next major release Geant4.9.0 is scheduled for end of June. A
Geant4 validation suite with 10000 jobs will start for about 3 weeks
on the GRID.
- Nightly builds for windows status: the complete software stack can
be build, what remains is the move of the build results to AFS space.
- Agreed to open a slot in the nightlies to exercise the compiler gcc
4.1. It may require to provide binaries for the externals for this
- The repository of the Grid client libraries in AFS is in place and
Stefan is working on a version of LCGCMT to make use on
these. LHCb will be testing the installation once the LCGCMT
package has been adapted.
- A number of new versions for externals has been requested (Boost,
Python, Qt4, kcachegrind, etc.)
- Presented plan for CORAL, POOL and COOL. Requested feedback from
- Preparation of LCG_51 configuration. New versions of frontier and
dcache being installed.
- New requests from CMS to ROOT for the schema evolution support and
the customization of the compression algorithms.
- Oliver Keeble reported on the progress made recently on the
installation of client libraries of GFAL and LFC for the SL4
architecture. A proposal with detailed information for doing this
installation would be circulated.
- Configuration LCG_50b with a number of bug fix releases will be
released, a list of proposed changes had been sent. The release date
of COOL 2.1.0 has been proposed for March 23.
- The availability of the client library for Dcache for SL4
needs to be checked as soon as possible.
- The client libraries for GFAL, LFC has been built for SLC4 but they
are not yet ready to be distributed and installed to the LCG external
- The priority for the common RFIO library between DPM and Castor
needs to be reviewed. ATLAS relies on it for sites with a
mixture of Castor and DPM.
- The first batch of tests for CORAL and POOL are now available as
part of the nightlies. The ROOT project will also provide tests
for the nightlies after the ROOT workshop.
- Configuration LCG_50a. This is a bug fix configuration based on
LCG_50 including bug fix releases of ROOT and CORAL, and some patches
on LCGCMT. It is being used by LHCb and ATLAS.
- Agreed in the installation of Java and associated libraries
required by ATLAS in the LCG externals. SPI is going to provide space
in AFS and guidelines for the installation.
- Line management changes: Pere Mato is replacing John Harvey as
PH/SFT group leader from the first of February. Andreas Pfeiffer
will be leaving the SFT group and joining the CMS experiment.
- The proposal that Stefan Roiser takes over from Andreas Pfeiffer as
SPI project leadership was agreed by the AF.
- Oliver Keeble from the LCG Grid Deployment (GD) area will be the
link-person between the AA and the GD with the main goal to improve
the communication between AA, GD and the experiments.
- Slow progress in the nightly build system concerning the
incorporation of CORAL/POOL tests and support for the Windows
- ATLAS reported the IO/Reflex problem in the in the Muon event model.
The problem has been fixed and a new patch of ROOT is going to be
- Problems has been reported with CERNLIB on SLC4 64-bit (by using the
h2root utility) but those has been solved.
- New externals:
- Agreed the installation of Qt4 for Windows into the LCG external
- No objections to install the fftw package and build the
corresponding ROOT plugin.
- New Python modules have requested: NumPy and MathPlotLib.
- Ian Bird and Les Robertson informed on request builds of GFAL, LFC
client libraries for SLC4. The problem (at least for GFAL and LFC) is
the external dependency on VDT. It is being discussed in the
WLCG workshop this week. No time estimate yet.
- Port to CMT of CORAL has been completed. The libraries and the
tests are built successfully with CMT every night, but the tests are
not yet adapted to QMtest.
- Agreed to use "dev1" slot for the preparation of LCG_50 and start
freezing the versions of ROOT, SEAL, CORAL, etc.
- Agreed to add an optimized platform in addition to the debug ones
in one of the slots.
- LCG_50. Target date for the release is the beginning of next
week. COOL and CORAL will be build with CMT and POOL with SCRAM.
- The nightly build system is currently operational, although not yet
complete, it is giving already good service. Two slots are
currently defined "dev" and "dev1". The two most urgent next steps are
the inclusion of the tests of CORAL/POOL as part of the nightly checks
and the Windows builds.
- Two new configurations in preparation LCG_49 and LCG_50. Both
include the latest patch from ROOT 5.14.00. LCG_50 will include in
addition COOL 2.0.0, which is being tested by LHCb and ATLAS, and new
a version of Frontier client.
- The SLC4 migration of the CERN/IT central services has been delayed
by 2 weeks. The builds of GFAL, LFC client libraries for SLC4 are not
existing and this will prevent to release any AA software component
that depend on them.
- Preliminary report by Philippe Charpentier on the outcome of Grid
client libraries and tools meeting that took place between some
members of AA, LHCb and Grid deployment with the idea to install all
the Grid client libraries and utilities on the LCG externals area as
it has been done with some of them already.
- The overall status of the AA nightly builds is that it is basically
is working. Gaudi has also been included.
- The basic testing of the ROOT production release at the level of
what can be done in the AA has taken place with the help of the
nightlies. Experiments has not done thorough validation due to several
- New production versions of ROOT and GEANT4 will be released as
scheduled this week.
- Discussed the proposal
for the migration of the SEAL functionality which is still left in the
project: component model, plugin management and foundation
libraries. The proposal was endorsed by the AF.
- A. Pfeiffer reported on the progress made in the nightly
build system and conversion to CMT. The system is basically ready
for the experiments to test.
The page http://lcgapp.cern.ch/spi/aaLibrarian/nightlies/index.htmlcontains
the information of the current nightlies and their status. The
binaries are installed in AFS at
- LHCb has triggered a exploratory meeting to understand the
interaction of AA and the middleware client software. P.
Charpentier will report on the next AF.
- Agreed to release configuration LCG_48a. This will include among
other small changes the ROOT (5.13.04c) and CORAL (1.6.2)
bug fix releases.
- Production versions of ROOT and Geant4 are scheduled for December
- Progress reported in the AA nightly build system. Experiments should
be able to test builds for some limited number of platforms in about
- The new lxplus/lxbatch service on SLC4 will be mainly based on
64bit architecture. The conclusion based on the tests that have been
performed by the experiments it that it shouldn't be a problem.
- The concerns on the distribution and deployment of middleware client
packages continue to be of actuality. Before moving the production to
SLC4, these issues will need to be sorted out. Philippe Charpentier
will be contacting Flavia Donno to start a discussion and prepare a
proposal on how to have a consistent client release of middleware
software for the experiments. Philippe will inform the AF of the
- Castor problems. The upgrade of the ATLAS and LHCb Castor
stagers had to be reverted in oder to allow old applications to
continue to run. A new version of the client library compatible with
the Castor 2.0 API is being prepared by the castor development team..
- The LCG_48 release, which includes SEAL 1.9.0, was almost
completed at the time of the meeting.
- Nightly builds are being developed for the AA projects. Andreas
described briefly the work he started to do to implement nightly
builds for the AA software. Small presentation is being prepared for
the next AA meeting.
- Enabling SS2 streaming instructions by default in AA code has been
put on hold due to the recent problems observed by LHCb when running
on the Grid (a sizable fraction of Pentium 3 nodes). It is needed to
get more feedback from the experiments before taking a decision.
- Status of LCG_47b and LCG_48 configurations
- Initial feedback from the LHCC reviewers quite positive towards the
Applications Area. Good progress was reported with no major
concerns having been identified. Action must be taken soon to ensure
the required level of manpower remains available beyond 2007. The
development of PROOF is progressing satisfactorily, but clear
decisions by the experiments are urged.
- Reported that the build of configuration LCG_47, which includes the
new development release of ROOT, is completed and will be
- Agreed to start preparing LCG_48 with new version of SEAL (1.9.0)
and ROOT (5.13/04) and with possible changes in COOL.
- ATLAS is actively looking at the persistence of the
event model in view to identify possible optimizations.
- Looking at the possibility to organize ROOT tutorials for data
analysis in a regular manner for experiments and CERN in general.
- General discussion about the preliminary outcomes from the AA
internal review. The projects will study carefully the final
recommendations when they will be ready. Preparation for the LCG
- New configuration LCG_47. It was agreed to wait maximum 24
hours to define the configuration in order to include the new Castor
client version. SEAL will go as is since the bugs in the
component model, mainly related to multi-threading, have not been been
fixed yet. The binaries for the compiler version gcc344 will not be
included for this configuration.
- Tag name for SLC4 with gcc 3.4.6 will have only the two most
significant version digits (e.g. slc4_ia32_gcc34) for all the new
builds and links will be created for the existing releases.
- AA internal review. Reviewed the available material
for the internal review.
- The MB agreed on 22 August to have a common RFIO library between DPM
and Castor in a time scale of mid-October. Meanwhile, it
was agreed in the AF in order to help CMS immediately, that ROOT will
be modified to provide two different and independent plugins: one for
RFIO, which should not include any Castor specific call, and
another one for Castor.
- A new ROOT development version (5.13.02) has been released today.
The agreement of the experiments was to wait for couple of weeks to
allow a new version of SEAL, which should include some bug fixes in
the area of thread safeness, to be ready.
- Agreed to install latest versions of Qt4 and pyQt together with a
new version of HippoDraw in the LCG external area on request of ATLAS.
- Agreed to request the deployment of the "CERN" patch to
correct the recently gcc compiler problem that was introduced with the
SLC3 updates while waiting for the official redhat one.
- Due to the recent change of the gcc version installed in the SCL4
systems, It was agreed to define the platform tag including only
the 2 more significant digits of the compiler version (e.g.
slc4_ia32_gcc34) and make a link on the existing packages that have
been released already.
- The testing
procedure put in place to validate the new production version of
the ROOT package before releasing it was followed and it was a half
success. All the AA projects could test their software but only half
of the experiments managed to do it in the timing and in the form that
the pre-releases were provided.
- ROOT 5.12.00 was released on July 11th as scheduled. The new
configuration LCG_46, which includes new ROOT, GSL, Castor
versions and builds of all AA projects, was ready in only 24 hours and
is being integrated by the experiments.
- The ROOT team have made very interesting I/O optimizations
(read ahead, request grouping, and so on) available in the
released version. These optimizations can not be implemented
when using plug-ins that provide posix like protocols such as rfio,
gfal, etc.. If changes in the API could be accommodated then
experiments could benefit from the optimizations. This issue will be
- Agreed to make a new configuration LCG_46a to include the new
version of dcap/dcache.
- Experiments are proceeding with the integration of LCG_46 in
their releases. No show stopper found so far.
- Agreed that the ROOT and POOL projects will add in their test suit
specific tests to verify that they can read and write files in the
mass storage system(s) using the provided protocols and plugins.
- ROOT version 5.12 will be released on July 11th. It was agreed to
make a complete new release of LCG (LCG_46) as soon as possible after
the ROOT release.
- CMS has serious problems with multi-threading in the area of Qt in
case of two concurrent event loops. Asking from the ROOT team to
understand the problem and find a solution.
- Geant4 8.1 being released. It comes with a new version of CLHEP. It
is the first release of Geant4 with the proper license (following the
- Report on the progress for the forthcoming complete LCG-AA release
(configuration LCG_45). New patch for ROOT was decided.CORAL
will come with new features in the ConnectionService for the
connection re-trial and fail-over cases.
- Geant4 started the full scale test for version 8.
- Discussion on the MacOSX platform. The majority of the external
libraries are already available. The AA projects need to start
building basically now.
- The EuroPython conference is going to be held at CERN this year from
3rd to 5th July. Contributions are welcome.
- The CMS request on handling the writing of SMatrix objects using
Double32_t will be dealt with the experts outside the AF.
- New version 5.11.06 of ROOT has released and new LCG_45
configuration is in preparation.
- Analysed the latest Castor incident caused by the deployment of the
latest release of the Castor libraries. The installation of the
Castor libraries in the external area should help in these cases.
- New requests from the experiments (CMS, ATLAS, LHCb) on various
subjects has been reported.
- Discussion on the possible workshop to discuss/define strategy for
interfacing CAD systems to our geometry models in view of the two
projects starting in this area. The conclusion has been not to
organize a formal workshop, however the projects will be contacted
in case they are interested to have a discussion with the people
having some experience on these issues.
- Decided to look into a possible migration of the projects using
SCRAM to the new version 1. The LCG/AA librarian, the release
managers of the projects affected (SEAL, POOL, CORAL, COOL, GENSER)
will get together with the CMS specialists to evaluate the work.
- Configurations LCG_43b and LCG_44 has been released.
- ROOT development release is scheduled for May 31. The production
version can either be done end of June or end of September. Input
from experiments is required.
- Presentation and discussion on the steps towards making ROOT
multithread safe. The I/O part could be implemented in a quite short
- Agreed to request the installation of Castor in the LCG external
area and add it as part of the LCG configuration.
- New ROOT development release 5.11/02. Rene presented the main
highlights for the release.
- New configuration will be prepared to include the new release of
- LHCb reported potential problems with authentication for COOL when
deployed at Tier1s as the current s/w is based on database
username/password in clear.
- The Oracle instant client library will be packaged in the LCG
externals. Oracle has agreed that it can be freely distributed for
LCG and be included in the experiment's software
- LCG_43 configuration is complete. It includes a new functional
release of COOL together with bug fix and rebuild releases for the
- Started to port projects to the new SLC4 platform (all externals,
SEAL and RELAX already done)
- For the Mac platform, the hardware architecture will need to
be included in the platform tag in addition to the OS and compiler
to take into account the new models. For example: osx104_ppc_gcc402,
- First discussion on the evolution of HepMC and in particular its
dependencies to external packages. No final conclusions but more
- Presentation of the [meta] plan for finalizing the migration of
SEAL functionality to ROOT. The presented plan was in general
- Status of the various releases in the pipeline needed by the
- The SLC4 certification has been declared finished. No
showstoppers. Now we will need to start building the AA software for
this new platform.
- Concluded on the policy change about C++ exceptions. The
note has been produced and accepted by the AF.
- New proposed dates for the AA internal review are 18-20 September.
- A proposal for categorizing the AA releases has been accepted.
There will be "rebuild releases", "bug fix releases" and "functional
- The releases of AA packages are progressing with minor delays with
respect to the schedule.
- Ratified the proposal to inherit the specific exception base
classes of each AA package (POOL, COOL, etc.) directly from
std::exception. Pere will send detailed information to final users.
- Tentative schedule for the next AA internal review is
beginning of October.
- ROOT release is scheduled now for next Thursday (2nd March). A new
build of the LCG/AA packages will be needed. Dirk will circulate to
the AF the plans for POOL, CORAL and COOL releases with the
functionality included and Andreas the proposal for the new LCG
- Discussion on the grid packages needed by the AA libraries and
their dependencies. The discussion was triggered by the need to
provide a GFAL installation in LCG externals.
- Status of the latest releases of AA software: migration to CORAL,
and Reflex in ROOT done. AMD64 port still under test in POOL and
- Added a permanent item in the AF meeting to discuss possible
topics for future AA meetings
- New quarterly report system for AA based mainly on milestones has
been put in place. The report for 2005Q4 is being finished.
- The installation of the version required by the experiments of
HepMC in the LCG external area as a standalone package is
- The next series of releases of LCG software (SEAL, POOL, CORAL,
COOL) are scheduled for the end of next week. They will complete the
migration from RAL to CORAL and from Reflex in SEAL to Reflex in
ROOT. These releases will also culminate the complete porting of all
the LCG (internal/external) software to gcc344 and the amd64
- Installations for HyperNews at CERN are available from the URL
http://hypernews.cern.ch. Presently the pilot service house two
active fora, one each for Atlas and CMS.
- The proposal for a testbed PROOF
system is being prepared in collaboration with Alice and CMS. The
main objectives were discussed.
- The production release of ROOT 5
is scheduled next Thursday. CMS requested that the complete LCG
software stack should released based on this ROOT version before
CERN closes. Releases of POOL and COOL based on CORAL and on
Reflex in ROOT will follow soon after.
- The port for SEAL to gcc344/amd64 was finished, version 1.7.7 has
been built with these new platforms. POOL/COOL should be making the
necessary changes to support it.
- Reviewed the preliminary conclusions and recommendations of the
LCG Comprehensive review for what concerns the Applications
Area. Preparation of a proposal together with Alice and CMS
for a PROOF testbed cluster with 150-200 nodes.
- The Linear Algebra package proposal was agreed by the AF. The new
package will be part of the production release of ROOT in December.
- CORAL 1.0.0 has been released. ATLAS is ready to try it.
Work has started in POOL and COOL to adapt to this new version of
the Relation Access Layer (RAL).
- The experiments (ATLAS, LHCb) are migrating to the new series of
releases and it going well in general. CMS is testing the
Frontier plugin and some problems are being found.
- The port of SEAL to gcc-344 ( ia32 and amd64) is basically done.
The idea is to start building the POOL products (POOL, COOL, CORAL)
for these new platforms.
- The proposal for HepMC package maintenance was agreed by the AF.
The details will be coordinated by Paolo Bartalini in the context of
the Generators project.
- The AF agreed to setup the HyperNews prototype pilot service
requested by ATLAS and CMS. The time scale will be to have the
installation by the end of next week.
- ROOT 5.06 was released two days ago. The binaries are already
built in LCG externals area (with the optional components such as
MathCore, MathMore, etc.). The final production release is scheduled
for 15th December.
- POOL has been tested successfully with ROOT5. This version is
interesting for experiments because of the Python interface based on
Reflex dictionaries. The complete suit of the LCG software should be
ready for the beginning of next week.
- Presented the proposed release schedule for the various products
of the POOL project
- CHEP paper submissions discussed.
- Clarification on the maintenance of pacman and usage by
experiments. Agreement that SPI provides tarball repository for all
packages (internal and external).
- Binaries for Fluka will be installed in the external area.
- Agenda for the AA session in the LCG comprehensive review (14-15
- Parallel validation releases for POOL with ROOT5 and CORAL.
- New requests from ATLAS: Reflex checksum utility and new platform
- Initial discussion on the next steps for PROOF. Agreed to have a
series of presentations to the experiments (in the format of
presentation/demo during the collaboration software week) to sell
the product and get what the real needs from the experiments are.
- Quick summary of the ROOT workshop that took place last week.
- FLUKA binary installation: the AF will coordinate a derogation
from the FLUKA licensing conditions for all LHC experiments.
- Progress in the validation by the experiments of the version of
POOL based on Reflex.
- SPI project. Andreas Pfeiffer will be the new SPI project leader
starting from the beginning of October.
- Following a discussion and some corrections to be implemented, the
AF endorses the AA Plan for Phase 2 document.
- ROOT development version 5.04/00 was released last Tuesday.
- POOL version 2.2.0 (first version of POOL based on Reflex) has
- The compiler gcc 3.4.3 (instead of the currently agreed gcc 3.4.4)
will be used for the AMD64 platform porting.
- The technical training offered by Apple should only be scheduled
in-time when the AA software is fully available for the Apple
- The AA plan exercise could be concluding now. The only pending
issue is the interest of the experiments (ATLAS, CMS, LHCb) to have
a Linear Algebra package independent of the ROOT framework. A
prototype development has been proposed. Minor changes will be done
in the plan document and feedback from experiment are expected in a
time scale of one week.
- Pacman. No immediate hope to make the product run natively on
Windows (is claimed to run well under cygwin).
- New releases. POOL will have two branches: one to validate Reflex,
and one to validate the new improved performance (CPU, I/O) still
based on Reflection.
ROOT 5 development release scheduled by 20th September (just before
the ROOT workshop)
- New version of Python (2.4) was decided for any new version of the
LCG software starting from beginning of December.
- List of the SEAL packages needed by ATLAS, CMS and LHCb and
- License issues for SEAL packages and LCG/AA in general. ROOT
copyright and license needed further clarification.
- Status of the various releases SEAL, POOL, ROOT, etc.
- Agreed on the evaluation of a possible bulletin broad service
(e.g. HyperNews) by SPI
2005 (special planning meeting)
- The following conclusions were
reached in the final wrap-up:
- The ROOT team will take the maintenance of existing SEAL
packages (this will be reflected in the plan i.e. the creation
of a work package ‘SEAL’ with a coordinator)
- The ROOT team should help the experiments and POOL on how to
adopt the new merged components.
- Linear Algebra. No conclusion for the proposal to have an
evaluation of an independent (from TObject) version of the
linear algebra package.
- Study the implications of packaging to understand how the
experiments can mix latest releases from certain packages to be
used in conjunction with the usually older, base/core packages.
- AA plans. In general the experiments were not
ready to discuss them. Proposed and agreed a special AF meeting
dedicated for that. The schedule for finalizing the plans will be
- Geant4 geometry I/O. Solution was
devised that will fulfill the needs of ROOT I/O and will add the
necessary protections to Geant4.
- New patch to ROOT 4.02/02 will be
required due to some failing regression tests in POOL.
- Finalization of the AA plans delayed two weeks.
- ROOT 5 development version released and including initial
versions of Mathcore, Reflex, Cintex and RooFit.
- MathCore proposal has been agreed. Agreement on
the proposed API. The library will be included in the end of June
release of ROOT. Independent tar file with MathCore will also be
made available at the same time.
- Frontier as a POOL/RAL plugin. Agreed to
develop an proof of concept prototype to evaluate performance on
read only databases.
- POOL 2.1 released. Coordination is need to
avoid problems with incoherent/incompatible grid middleware
- Updated list of SEAL+ROOT merge topics
- Python bindings for Reflex: decision to pursuit PyROOT in tandem
with Cintex for the short term
- GCC 3.4.4 to be used for the next iteration with the amd64 and
- LCG software will not be build anymore for RH73 version of the
- Various new versions of POOL, ROOT, Geant4 expected at the end on
- Status of the writing of the AA chapter in the LCG TRD.
- Decided that PROOF will be a work package in the CORE (ROOT)
- Report on the current functionality of PyReflex. Need for a
solution in a short time scale.
- Decided to build a new series of versions of SEAL, POOL, etc,
based on the ROOT production version 4.04/02
- The current level of backward compatibility for the ROOT
backend of the POOL version with Reflex adequate to
- Agreed that the new "toolbox" configurations will be take the new
MacOSX version into account.
- The final report of the AA internal review has been very well
received by the project leaders.
- Each project (SPI, ROOT, POOL, SIMU) will prepare a work plan
taking into account the review recommendations
- Established the process for handling the details in the SEAL+ROOT
merge. The relative priorities for the first set of topics has been
assigned. It is understood that work items of the "old" plans have
- The AA priority for next supported platform is still AMD64 with
- Python courses will continue to be organized by the CERN technical
- Started to have discussions on the relative priorities of the list
of topics for the SEAL+ROOT merge
- POOL should be planning for the first release compatible with
Reflex by end of June.
- Problem on the GNU Fortran compiler:
- What is the current usage of Fortran in the experiments
- This is a real problem. Fortran will probably be use for
- CERN/IT will take an active role in finding a solution.
- AA internal review follow-up. Detailed plans for most of the work
items are still needed. It was agreed that a list with all the
topics will be produced together with all the relevant issues.
- Announced new ROOT development release
- External libraries. A set of new additions are being requested by
- It was agreed to interface ROOT to SRM. This will be done as a
- RooFit will be packaged and re-distributed within ROOT starting
from version 5
- Encountered problems with schema evolution between ROOT3 and ROOT4
based POOL versions. Problems are fixed and new releases planned.
- Porting to gcc 3.4.3 compiler: reporting progress,
installation issues, etc.
- AFS volumes containing the releases of AA software and external
libraries made read only to improve performance.
- Agreed to have the start of the AA meetings to 16:30 and with a
cycle of every two weeks.
- The AA internal review agenda is being prepared in http://agenda.cern.ch/fullAgenda.php?ida=a051527
- Tentatively we could propose to stop building and testing new AA
software for RH 7.3 starting in July
- Vincenzo will act as champion and driver for the AMD64/gcc 3.4.3
- ATLAS suggests a review of CLHEP usage is needed in view of the
persistency problems seen and the CLHEP2 migration workload to come
- AA will (with the help of the experiments) maintain a list of AA
software versions/platforms currently in use in the experiments
- It was agreed that RH7.3/gcc 3.2 can be retired on a summer
timescale, when ATLAS ceases builds for this configuration
- the gcc 3.4.3 port will be done for both Intel 32-bit and AMD64
- LCG AA will ask the CLHEP project to regard schema evolution
compatibility as a constraining requirement in CLHEP code changes
- LCG AA will include CLHEP tests in the POOL regression test suite
to check against new POOL/ROOT and CLHEP versions
- SLC3 sysadmins have been asked to keep in place the runtime
workaround that enables POOL 1.7 operation on SLC3
- RAL will be partitioned in place in POOL to be independent of the
rest of POOL
- AMD64/gcc 3.4.3 is agreed as the next platform priority [ATLAS
adds that ATLAS TDAQ requests 32-bit gcc 3.4.3 support also.]
- SPI proposals for bugfix policy and cleanup of release policy were
revised and agreed
- The following points will be raised with the PEB:
- there should be a discussion on the application of xrootd in
the LHC experiments, whether in the AF or some other forum. With
xrootd at the border between applications and gridland, the
appropriate forum is not clear given present scopes.
- serious problems with RFIO (described in an appendix to the
minutes) are outstanding and need to be addressed if RFIO is to
continue to be used.
- the suggestion should be put to the experiments that if xrootd
is proven to be a fully functional replacement for RFIO, it
should be adopted as the RFIO replacement. Rene said xrootd
covers the functionality of RFIO and adds others like load
- Frederic Hemmer joined the meeting for a discussion of middleware
- After discussion with Frederic, AF reaffirms the proposals that
- SPI's librarians integration meeting be used as a forum in
which gLite and GDA join the current participants in a managed
process of proposing and agreeing on packages/versions
- gLite (and GDA) make use of SPI's external software services
as an aid to coherent management of package/ version usage
- the AF notes (and regrets) the lack of a forum to effectively
address data cataloging and management plans in the experiments and
identify commonalities among them
- SEAL has built with ROOT 4.02/00; POOL will build a prerelease
with it next week. POOL will then need validation feedback from the
experiments. Also would like tests that can be integrated in POOL
- SEAL partitioning is underway (with much remaining to be done)
- Cintex is ready for experiment trials with their EDM classes
- AttributeList, RAL will move from POOL to SEAL, possibly as early
as the next SEAL internal release.
- full SLC3 build of AA software and externals checks out OK and
migration will now be done (has since been done).
- POOL supports SLC3 from release 1.8 onwards.
- POOL2 production release is expected in late January, following
the (now imminent) ROOT production release and experiment
- POOL does not support persistifying boost. Any persistifying of
boost classes is at the user's own risk; POOL will not invest effort
in fixing schema evolution problems arising from boost changes.
- SEAL partitioning needs to be given higher priority.
- Rene has reported a ROOT fix (available in ROOT 4.01/04) for the
CLHEP namespace schema change problem.
- Boost smart pointers will not be used on public interfaces.
- An experiment request to introduce AMD64/gcc platform support is
likely before long.
- AA will share the Wed AA meeting slot with ARDA, as requested by
- SPI will request dedicated lxbuild servers for AA
- A need from POOL for Globus file catalog external library support
- A Middleware visit to the AF will be planned to discuss this and
other AA-Middleware issues
- planning is actively underway for transferring AA leadership to a
new project leader in Feb 2005. AA activities will continue to be
part of the LCG project. The leadership change and LCG Phase 1/2
transition will be an occasion for examining potential changes to
- Ian Bird attended to discuss AA/GDA issues, summarized as follows:
- AF proposes that gLite (and GDA) make use of SPI's external
software services as an aid to coherent management of package/
- AF proposes that SPI's librarians integration meeting be used
as a forum in which gLite and GDA join the current participants
in a managed process of proposing and agreeing on
- GDA will begin in-house POOL testing using the POOL test
suite; the POOL team will give any needed assistance in setup
- The DC experience with EDG RLS led to a reimplementation of
the RLS back end by GDA to address its deficiencies. The AF
would like to understand how this experience and work is being
used by gLite.
- Ian made the clear and fundamental statement that with present
web services technology, it is not possible to achieve the
performance required of grid services. Why then are we using
- Ian informed us that lxplus machines have a grid UI
environment available, but this seems too little known.
- Large numbers of people will need to have grid certificates if
grid enabled services successfully play a role in analysis
contexts which are ramping up now. Is there a plan for handling
this? Is the FNAL Kerberos solution an attractive option that is
being looked at?
- The issue of SLC3/RH Enterprise runtime (non)compatibility was
raised again. This is a problem for the AA but not something
that can be resolved by the AA. How do we get it resolved.
- AF agreed that the experiments are advanced enough on SLC3
migration that it is OK for grid sites to start moving to SLC3.
- Dirk needs DB applications input from the experiments for the 3D
project, in addition to data volume input.
- POOL has an agreed plan for enabling POOL support within the ROOT
environment. The focus has been on 1) providing a POOL plugin for
ROOT to enable support for POOL references in ROOT, and 2) using
Reflex to make LCG dictionary based classes accessible in ROOT. If
there are requests beyond what is presently planned, e.g. support
for writing POOL files in the ROOT environment, the experiments need
to make them clear.
- US CMS has implemented a Globus file catalog back end for the POOL
file catalog interface. It will be added to POOL when complete and
- Fermilab people have implemented a first version of a Frontier
back end to the POOL storage manager.
- a serious compatibility problem has come to light for POOL
persistency of CLHEP classes. CLHEP 1.9 introduces a CLHEP
namespace, which will render CLHEP objects written with POOL in
CLHEP 1.8 based applications unreadable from CLHEP 1.9 based
applications. Possible solutions are being investigated. This will
not delay, but rather accelerate, the introduction of a CLHEP
- the channel for proposals for new packages and versions will be
SPI's librarians integration meeting (LIM), with the AF used for
- suggestions for Savannah improvements are invited; the developer
will be visiting SPI soon.
- Geant4 will use CLHEP 1.9 in Geant4 7.0, not CLHEP 2.0 as planned
- POOL 2 release expected in ~2 weeks, using ROOT4 development
version with new persistency, and automatic schema evolution, for
most STL containers.
- Production version will come out at end of year, when the next
ROOT production release is scheduled.
- This plan is compatible (just!) with CMS and ATLAS schedules; they
both require a production version no later than mid January.
- The experiments emphasised that ABI compatibility between SLC3 and
RH Enterprise Linux is an essential requirement.
- Experiment manpower commitments to POOL/CondDB were discussed.
Many still have to be firmed up, and it is important that once firm
they are also stable.
- There is interest in CMS and other quarters in proceeding promptly
with implementing a Reflex back end that fills CINT data structures.
Identifying available people with the expertise to work on this is a
- The ROOT file format for the POOL2 prereleases is not guaranteed
to be stable. When POOL2 is released, it will include a
specification of the file format stability commitments.
- ATLAS, LHCb are asked to test the POOL2 prereleases now.
- A collaboration between Fermilab and the 3D project on FroNtier, a
high performance and scalable multi-tier database access system,
including POOL integration, looks very promising.
- A report to the Desktop Forum on MacOSX as a development platform
has been prepared.
- July 29,
- Geant4 expects to support CLHEP 1.9.x in a G4 development version
by end September.
- Andreas Pfeiffer will initiate a bi-weekly Librarians Integration
Meeting (LIM) to discuss configuration/integration of AA software
into the experiments, and coordination of AA (pre-)releases with the
- SPI will provide a "single line command" for source download and
build for AA software, with the first version available in
- SPI is developing an XML-based build configuration description
which will be the basis for preparing configuration information used
by SCRAM, CMT and autoconf/gmake.
- SPI will implement autoconf/gmake based builds, using the XML
based configuration information.
- External packages are now built for SLC3 and MacOSX.
- Alberto flags the need for more Windows expertise if we are to
support Windows at the same level as the other platforms. Are there
Windows experts out there who can help?
- A means of coordinating external package versions between AA, GDA
and experiments is needed.
- July 15,
- The AF supports the selection of Gabriele Cosmo to lead the
- SLC3 platform names will follow a slc3_ia32_gcc323 pattern
- External packages for SLC3 are about 80-90% done
- External packages for MacOSX are at a similar level of completion
- icc8 porting will continue. About 50% of external packages are
done. Priority lower than SLC3 and MacOSX.
- July 1,
- GDML as exchange format is progressing with much remaining to do.
AF supports going forward, with immediate goal to demonstrate Geant4
write/read capability on complex (LHC detector) geometries.
- ROOT4 production release will be next week, with SEAL and POOL
releases using it following closely.
- eliminating SEAL's ROOT dependence will be investigated, in the
context of beginning to split SEAL into independently releasable
- POOL collections WP meetings have been regularized on a biweekly
cycle, with progress and discussions on several fronts.
- IT/DB now has an Oracle client distribution page.
- June 10,
- Starting with SEAL 1.4 and POOL 1.7, ROOT4 based builds as well as
ROOT3 will be provided on the supported platforms. Experiments are
encouraged to start evaluating and validating ROOT4. ROOT3 based
versions will continue to be provided until the experiments fully
certify ROOT4 and decide to drop ROOT3.
- The experiments should consider use of ROOT4 based POOL to be 'at
their own risk' until they have performed adequate validation to
certify it for their own production use.
- Markus and Rene believe that backward compatibility issues have
been addressed and POOL/ROOT4 will support reading of ROOT3 based
files and appending to ROOT3 based files. It will be for POOL and
the experiments to validate this.
- The POOL team needs more real-world POOL tests from the
experiments to integrate into their test suite.
- Rene gave a presentation describing ROOT's plans for expanded
RDBMS support, in particular mapping TTrees to relational tables via
an RDBMS extension of the standard ROOT I/O engine.
- There were reports on CondDB and simu project developments.
- May 13,
- Alberto will take up with Andreas moving forward on the appwork
evaluation (the gmake+autoconf build system).
- Ian Bird attended this meeting for a discussion of grid deployment
area liaison issues.
- GDA will integrate AA-provided test suites to validate AA-GDA
software integration (e.g. POOL-RLS) into the test suite used for
deployment validation, in order to expose integration problems
- The packaging tool for the coming new middleware has still to be
decided. There have been discussions with Miron Livny about his
group doing this, using pacman. Pacman is what SPI has been
investigating and we will continue with that since it seems to be
the most likely common solution.
- AA needs a good communication channel to GDA for e.g. discussion
of release schedules and release announcements.
- We will arrange for a grid deployment area person to attend an AF
meeting dedicated to GDA liaison issues every few months.
- April 29,
- Discussion on MacOSX in the experiments. There will be an AAM
meeting on Mac computing soon. Planning porting work, and its
relative priority, is underway.
- boost compilation problems with icc8.
- AF will meet with GDA (Ian Bird) soon to discuss mutual issues.
- Conditions DB (sub)project is in urgent need of specific
commitments on participation.
- April 15,
- The AF identified a candidate new level 1 AA milestone, an end
2004 milestone for the validation of Geant4 in production in this
year's ATLAS, CMS and LHCb data challenges. [Subsequently agreed by
- A plan will be developed for the gradual integration of ROOT4 with
AA software, with initial adoption fairly soon after its June
release, and usage of new functionality in ROOT4 for schema
evolution and improved STL persistency following later on timescales
consistent with backward compatibility requirements from the
- External and AA software have been ported to the CEL3 CERN
Enterprise Linux certification platform.
- April 1,
- Factor two speed-up on CMS POOL data store in production after
CMS/POOL/ROOT examination (optimized ROOT key access)
- LHCb pre-production with POOL should start next week
- ATLAS POOL problems (CLHEP matrix, vector) solved
[post-AF developments also]
- POOL workplan approved with mostly minor revisions
- Less minor: agreement to remove collection catalogs, as not within
- CEL3/gcc 3.2.3 external package builds completed; SEAL and POOL
- MacOSX accepted as a development platform and replacement for
Solaris as a ttest-build platform
- March 4,
- The experiment Data Challenges this year (ALICE excepted) will use
the present ROOT3 based POOL. A new version 4 of ROOT will be
released in a production version circa June. It has features
attractive to POOL such as improved STL container persistency. POOL
will develop a ROOT4 migration plan, with concurrent support for
ROOT3 and ROOT4 based POOL I/O a requirement after ROOT4 adoption in
order to meet the experiment requirements for 18 month readability
- The SEAL workplan was approved.
- The POOL workplan draft was discussed and some changes agreed
upon; with these changes the plan is acceptable. Dirk will update
the plan accordingly.
- Proposed new Level 1 (LHCC) milestones were discussed.
- ATLAS reports their simulation validation for Geant4 simulation
production is complete, with 500k events simulated and written to
- The substance of the SEAL workplan was agreed, in particular
including the new item discussed, mathlib:
- A broadened mathlib technical discussion forum including experts
from all experiments will be charged to develop a detailed technical
plan. See the minutes for more info.
- Pere will finish up the SEAL workplan document with the objective
of approving it at the next AF and PEB meetings. Next AF meeting
will address the POOL workplan.
- gcc 3.2.3 will be the next production compiler supported by LCG
AA. Timescale probably circa summer.
- Experiments are working on POOL performance/file size and making
substantial improvements by optimizing how they use POOL. It is
important that we capture the lessons in the POOL documentation.
- Dirk reported on problems with RLS and its interaction with POOL
that have seriously impacted CMS attempts to get production going on
- Experiments will consult directly with POOL, SEAL to ensure their
requirements are reflected in the draft workplans before we address
the plans in the AF.
- A mathlib project proposal based on the ROOT-SEAL discussions so
far will be available shortly.
- gcc 3.2.3 will be the next compiler supported by LCG AA. Timescale
is uncertain; possibly circa summer 2004. SPI will start preparing
for it on a time-available basis.
- icc version 8 is out and will be installed by SPI.
- SPI will provide a standard installation of CERNLIB (the kernel,
not satellite packages).
- A proposal is being prepared on how to deal with NAG-C and Iris
Explorer/Open Inventor licenses
- Discussion on need for POOL relational back end
- LHCb report on their almost-completed POOL integration milestone
- SEAL has distributed a preliminary workplan. It is at http://lcgapp.cern.ch/project/mgmt/seal-2004-draft.doc
- CMS requests item-by-item experiment input on interest and usage
plans for workplan items. Agreement is that experiments will
annotate draft workplans with their input prior to settling the
plans in the AF.
- LCG policies developed by SPI ratified.
- Problems with the icc compiler (not with code) are impeding the
completion of the icc port. Because icc is only a test platform,
this delay will not be allowed to delay the forthcoming SEAL and
- The forthcoming POOL release (1.4) will support Windows.
- Experiments are asked to present proposed new
integration/validation milestones (relating to the apps area) in the
- The ARDA kickoff workshop is delayed to January [21-22] because of
the difficulty of finding workable dates.
- Internal review report received. Constructive and useful. Project
leaders will write up first responses for next week.
- We will take up next week (we can hope for the last time this
cycle) LCG policies from SPI for ratification, after we receive an
iteration from Alberto reflecting recent input.
- SEAL will coordinate migration to CLHEP 2.0. Fix for CLHEP matrix
dictionary generation problem is in progress.
- appwork supports complete builds of POOL and SEAL and the projects
are asked to take a serious look at it. Next steps are remote sites
(done) and other platforms (icc, then Windows).
- Boost (not SWIG) will be our standard means of interfacing C++ to
Python, complemented by PyLCGDict.
- We agree we will work now to converge upon a single dictionary
(LCG and ROOT/CINT) and we will develop a concrete plan for this.
Discussions are underway between SEAL and ROOT.
- ARDA planning is starting and a kickoff workshop is being planned
(likely date has changed to the week of Jan 12)
- This was a special meeting on a proposal from Rene Brun for a
tighter collaboration between LCG AA and ROOT. No attempt to
summarize. Read the slides and notes. Will be the subject of an apps
area meeting on Dec 10.
- Ratification of LCG AA policies will be taken up at the next
regular AF meeting.
- SWIG vs. Boost as the mechanism for interfacing C++ classes to
Python is leaning the Boost way and will be taken up at the next
regular AF meeting with all experiments present.
- There will be (was) a special AF meeting on Oct 30 to hear a
proposal from Rene Brun on how to go forward with a tighter
collaboration between the applications area and ROOT (was very
constructive and will be reported out soon.)
- October 9,
- Statement on GSL and NAG expected within two weeks.
- First phase of SPI-G4 collaboration on infrastructure established
(involving Savannah and testing).
- Long and productive discussion on the simulation project. See the
minutes. Some of the main points: Generic framework's immediate
target is the physics validation project, which could make good use
of it immediately. Near term approach will be to use FLUGG to
integrate FLUKA as a simulation engine using the G4 geometry. Longer
term approach, from early next year when the ALICE/ROOT VMC using
the ROOT geometry model is ready, will be to use an XML exchange
format building on GDML work to transmit G4 geometries to the VMC.
The AA will make a proposal to SC2 to bring this persistent exchange
tool building on GDML within project scope.
- POOL and SEAL will put out releases ASAP to pick up ROOT 3.10
- We will designate certain specific releases as test build
releases. Only for these releases will all the external packages be
built for the test platforms.
- Build system: LHCb has a CMT-based SEAL Windows build working.
Regarding the longer term (cf last AF), a fast autoconf+make based
prototype is in place which builds POOL. It will be made public
- October 2,
- POOL needs more complex schema from the experiments to improve
- Mathlib workplan will come as part of an updated SEAL workplan in
the next quarter.
- Prototype for POOL RDBMS independence layer (file catalog and
collections) should be out by the internal review. Using OTL.
- Conditions DB work package needs to get going quickly.
- All agree on an objective of stabilizing external software
versions as our software development stabilizes. Goal will be ~6mo
typical lifetime for a given package version.
- Progress on the needed improvements to the SCRAM build system
identified in last year's evaluation has been disappointing. Most
seriously, support for Windows binary builds required by LHCb
remains incomplete. LHCb will choose an expedient approach to
implementing an interim solution. For the longer term we will
examine going to a conventional (gmake+autoconf) open software
approach. The matter will be addressed in the internal review.
- Discussion of generator services and FLUKA subprojects of the
- First prototype of a relational back end for the POOL object store
is emerging. CMS have expressed their need for it for conditions
data; other experiments are asked for input.
- ATLAS wants OSX support. CMS wants it also. Will not get any
official attention until the present set of additional compilers is
properly dealt with -- Windows, icc, ecc.
- July 24,
- Distribution and installation system for LCG and external software
is completed and under test. Information is on the SPI web.
- Discussion on simulation
- Planning for the applications area internal review in October
- July 10,
- A special meeting to address issues surrounding the PI work
program, the role of AIDA, and the integration of ROOT as an
analysis tool in the LCG software environment. No attempt to
summarize; read the full text.
- July 3,
- SPI's work on icc and ecc is complete except for MySQL related
issues. The ball is now in SEAL's court, followed by POOL. The MySQL
file catalog implementation will not be built on icc, ecc until the
MySQL issues are resolved; other implementations will be used.
- SPI is proceeding with Windows, where there is real progress. SEAL
now builds with SCRAM under Windows/cygwin using VC++. POOL is in
progress. SPI is making progress on building VS project files in
- POOL and SEAL will make it a target for one of the next releases
to agree on release procedures.
- A 2-3 day internal technical review of LCG software is planned for
the fall, tentatively for the last week of October.
- Software packaging and distribution service is urgently needed
- June 12,
- POOL/SEAL integration/acceptance milestones from ATLAS and CMS are
needed ASAP. We should have next-quarter milestones for this in the
coming quarterly report.
- Software packaging and distribution plan and service is urgently
needed from SPI.
- There continues to be little feedback on AIDA - the current
interface and the proposed changes from PI - from experiments or the
- Derek Feichtinger and Frederik Orellana will work on two tasks in
the distributed analysis area: 1) developing a remote launch service
for interactive analysis, to be used by PROOF and Ganga; 2) bringing
up Mannot (Ticino, Switzerland) as an operational LCG-1 production
- Technical design reviews for POOL and SEAL will be planned for
- May 22,
- User feedback is needed from the experiments for the AIDA
- Projects are asked to work with the experiments to establish
milestones for the integration and usage of LCG software in the
- Good progress in SPI on icc build of the software library. ecc
also in progress. gcc 3.2.3 will follow.
- We will move to ROOT 3.05.05 in the next POOL, SEAL releases.
- May 8,
- We are asking the experiments, the projects, ROOT, the apps area
in general for proposals as to how to employ the presently available
effort in distributed analysis.
- CMS is planning to use POOL/SEAL from June/July; see Vincenzo's
May 14 applications area meeting report
- The installation
layout policy proposed by SPI is adopted
- All libraries will include a prefix in the name (TBD, e.g. 'lcg_')
to remove the possibility of name clashes with libraries outside the
- A policy for version migration of external packages is needed
- The proposal to bring conditions DB
work into scope will be sent to the SC2
- April 17,
- IT will implement a CVS service acceptable to LCG-Apps in a few
months and we will use it.
- POOL, SEAL schedules are acceptably synchronized
- For the first test compiler, icc, SPI should be ready by end May,
SEAL by end June, POOL in their following release
- Vincenzo will do initial ROOT implementation of AIDA basics;
target is June.
- The POOL/SPI proposal on testing will become policy; give feedback
- We need a MySQL service. The experiments will present the issue to
- It is too early to give serious attention to a common XTI based
dictionary, but we should keep it in mind as an objective and plan
- March 13,
- The experiments need a working, supported implementation of AIDA
histogramming. This will be a ROOT based implementation. We are
trying to find someone to work on this.
- The standards and policies that define 'SPI compliance' must be
clearly agreed, documented and posted by SPI as LCG policies.
- Proper documentation is a release prerequisite. A package cannot
be tagged for inclusion in a release unless and until documentation
requirements are met.
- SPI will propose concrete and practical documentation requirements
that will provide a clear standard.
- A proposal to SC2 to bring common conditions DB work within apps
area scope will be prepared.
- Licensing. The apps area must have a GPL-type license in order to
function, because only GPL licensed software can use other GPL
software, and because we depend on many GPL packages. The PEB will
work to establish such a license as the LCG license.
- The PI project will develop a user friendly end user interface for
AIDA based on value semantics rather than pointers over the next few
weeks, for user evaluation.
- Simulation project setup status was summarized. It is coming
together quite well. Presentations to management and to the
applications area will take place over the next two weeks. An
initial workplan will be presented to the SC2 on March 14.
- CLHEP has accepted the LCG/SEAL proposal to use LCG infrastructure
to 'host' CLHEP (CVS repository, Savannah portal, other SPI
- LCG CLHEP activity will be organized as part of the SEAL/math
libraries work package
- Red Hat 7.3/gcc 3.2 will continue to be the sole release platform
for the moment, based on the needs of the experiments. Debug and
optimize versions will be provided.
- Builds for the purposes of testing will be set up for icc, ecc,
Visual C++ and Solaris CC in that priority order.
- The ROOT course was very well received. A further ROOT course, and
other courses (e.g. SCRAM) will be arranged.
- CVS structure and include convention are settled. See SPI web.
- ATLAS will pursue the 'gcc-xml path' from C++ header file object
descriptions through gcc-xml and the LCG dictionary to the ROOT
dictionary and ROOT I/O in POOL.
- We will propose to CLHEP that the LCG apps area host the CLHEP
software (CVS repository, Savannah portal, etc.) [proposal was
setup status was outlined.
- We are a C++ based project; core AA software will be in C++ at the
- Java based tools which may use parts of the core C++ software but
are not themselves used by the core software, are allowed as
- ROOT dictionary - LCG dictionary cross-population
- ROOT/CINT/LCG collaboration on common dictionary
- New projects: core libraries and services, physics interfaces
- Non-conclusive SCRAM/CMT discussion
- AF meetings were on hiatus while the architects were meeting in the
architecture blueprint RTAG Jun-Oct
- June 7,
- LCG dictionary following Gaudi/Athena dictionary interface (based
on Java reflection API)
- Population of the ROOT dictionary from the LCG dictionary
- May 15, 2002
- POOL will use a globally unique ID to identify files
- Deep copies need to be supported in POOL
- Savannah will be evaluated
- Apr 24, 2002
- Addressed persistency framework issues
- Component breakdown suggested by RTAG accepted as basis for work
- MySQL will be the relational technology for the initial prototype
- Jobs do not share files on write; a given file has no more than
one writer at a time
- Will support recovery at finer granularity than file level