AFmeeting 13 November 2008

Attendees

Agenda

• Line management issues
• Action Items
• Status of releases
• Platform and external libraries issues
• Experiments feedback
• R&D project status
• Topics for AA meetings
• AoB

version 1.0 (14/11/08 @ 16:00)

Line management issues (Pere)

• No news yet from the Fellow and Associates committee. The folders for the Technical and Doctoral students are available.
• Collected the abstracts for the CHEP papers proposed by the SFT group. Pere reminded that each 'submitter' is responsible of their submission.

Action items

• Links to R&D project in AA and SFT web pages (Pere)
• Not yet done. In this context, the evaluation of the DRUPAL content management system will be also done for SPI in addition to ongoing migration that is taking place for the ROOT web.

Status of releases

• LCG_55b (de-sealed)
• The configuration was finalized at the end of last week and announced on Monday.
• Pere mentioned that there has been basically a week delay, but this was very acceptable for the experiments.
• Marco remarked that it is important to decide well in advance the contents of the configuration in terms of versions of packages and supported platforms. This is give sufficient time for developers to adapt their code to the new environment.
• Pere explained what should be the normal modus of operandi for the releases of new configurations in the current phase of the projects. This is:
• Bug fixes and new functionality is made available in the nighlies as soon as the code is ready.
• Projects make sure that nighlies are always on 'green' for all platforms.
• Releases can be cut 'on demand' of the experiments with relatively short notice. It should take of the order of a couple of days.
• There was basically agreement with this with some some little concerns from Andrea in case large changes or additions are needed for a project, which prevents to have everything working within a day or two.
• LCG_56
• The objectives for this configuration are:
• Move to new ROOT pre-production version (5.21.06)
• Move to new Boost version 1.37 [originally 1.36 but decided on 1.37 at the end of the meeting since this version is already available and supports 'better' gcc-4.6  and VC-9]
• Move to new CMT v1r20p2008x. Wait for an official release in the next days/weeks. Consult with Gregrori. --> action Stefan.
• Platforms: SLC4/SLC5/gcc-4.3.2/gcc-3.4.6 (32/64-bits), MacOSX-10.5/gcc-4.0.1, Windows/VC++ 9.0/VC++ 7.1
• Introduce the new LCGCMT tags
• A pre-release should be made available towards the end of November and the production quality one mid-January.
• It was also mentioned that we need to prioritize on the platforms. For example in case VC9 is not ready by mid-January then we should still release for the other platforms.
• ROOT pre-production version expected by November 26. The list of changes is agreed. Rene mentioned some pending problems with STL/Windows/VC9. Huge memory leaks has been reported by Chris Jone (CMS). Still not yet decided if Reflex should have the new API and the fast look-up (x20).
• Concerning the new Boost version, it was decided to switch the nightly slot to start using it and see what breaks, and then  fix it. Liz reported that this new version has same changes in the interface of the Boost Thread library. 
• Pere mentioned the fact that GAUDI_ATLAS is always red. David will follow this with Charles but looking at it it resulted to be related to the inclusion of the package GaudiPoolDb, which is not used by ATLAS.

Platform and External Libraries issues

• Issues with SLC5
• Pere made a presentation with the decided SLC5 migration strategy and the status of testing SLC5 by AA and experiments in the LCG Management Board. The only concern was from Grid Deployment (Markus Schulz) about the availability of the client libraries for gcc-4.3 that will have to be done by hand (since Ethics does not support any other compiler than the native one). I can take up to 2 or 3 weeks.
• Phase 1 issues: Memory leaks ATLAS and OpenAFS
• David reported that they didn't run any more tests in the last 10 days because their tests were crashing the SLC5  machine with a "removed-link file" error due to an OpenAFS problem. They are waiting for somebody in IT-FIO are is handling the case. They also can not say anything in addition about the memory leak observation. They need to revisit once the problem is fixed on the AFS side. 
• CMS have been quiet successful on their compatibility tests. They have also build all the externals with gcc-4.3 and reported some small chnages needed in the thread library.
• Stefab mentioned the the current SLC5 installation misses some library required by the md5 Python module. 
• SLC5 should be available in the nightlies from now.
• Progress on Windows VC++ 9.0
• Status is the basically same, expect that Rene reported that the need of special macros of the STL library implementation are not necessary anymore.

Experiments feedback

• ATLAS
• Nothing additional to report.
• CMS
• Starting use some of the new things from the next LCG release (e.g. new ROOT). Too early for any feedback.
• Alice
• Nothing additional to report.
• LHCb
• Nothing additional to report.
  

R&D project status

• Multi-core project (Vincenzo)
• Presented the status to the PH management this week. On the discussion of the budget for next year, it is envisage to move material budget to hire more Fellows.
• Started with Fons to produce a kernel version to include the  KSM memory module that should help in recognizing virtual memory pages that are duplicated and that can point to the same physical one. Vincenzo will try it to see if the kernel is installable and eventually can run any of our code.
• Proposal to buy advanced hardware (with 6-core CPUs) that should be in the leading edge for the next 12 months. Fons will see if we can get a good deal with the hope to get it before the end of the year. 
• Virtualization (Pedrag)
• CernVM version 1 was released. So far so good. Only a coupe of minor problems reported. 120 downloads of this version so far. 
• Presentations in ACAT workshop and ATLAS software week.
• Similar presentation to the PH management.
• Jakob Blomer will be starting from the first of January as Doctoral student.

Topics for future AA meetings

• Nov 12: Eventually ACAT highlights. Lorenzo and Fons have agreed. 
• Early December we should have a presentation with the ROOT-LLVM proposal. Discussion and eventual decision will be taken in the context of the AF.

AoB

• Observed that Boost 1.37 was out. This has been tried by CMS. It was decided to use this version for the next configuration.
• There has been a request to introduce as a new external the "Oxbridge MT2 / Stransverse Mass Library", which is already used in some place in ATLAS. It was decided that these kind of requests should be channeled through the experiments, and not by the authors directly.
• Next AF meeting in two weeks.