-Welcome to summer student with Ilka
Roel Aaij (Ru Ai) - summer student working on the GUI
-Bastien is progressing quickly with the spider plots (or like). He will talk at the next meeting.
-PHYSTAT next week at CERN, see: http://phystat-lhc.web.cern.ch/phystat-lhc/
-Finance/Council meeting this week. Rumors are positive
Quoting a mail from Pere
"It seems that FC has approved the extra resources for the white paper.
This means that SFT will be able to ensure some continuity. The draft figure
that Jean-Jacques is talking about is 10 FTE for SFT operations for the
years 2008-9-10 of a total of 117 FTE for the department. In addition to
the operation, there is the possibility of extra resources (manpower and
material) for R&D for a total of 33 FTE for the department. This is for
the 3 projects (multi-core, virtualization, distribution) that we proposed
last year."
Meeting with CMS yesyerday to discuss
-Automatic Schema Evolution requirements
-I/O speed improvements (iterating today: they have trees with more than 2000 branches!)
see minutes below
The discussion started from the following documents:
proposal by Chris:
https://twiki.cern.ch/twiki/bin/view/CMS/FwEdmSchemaEvolutionProposal
prototype by Vincenzo:
https://twiki.cern.ch/twiki/bin/view/CMS/VISchemaEvolutionPrototype
René evaluated CMS proposal and reported that all the required
schema evolution cases are currently supported by ROOT. The most
generic schema evolution requires, anyway, to define a custom streamer
for the object types to be evolved. The following source code was
brought as example:
http://root.cern.ch/viewcvs/hist/src/TAxis.cxx?rev=1.83&content-type=text/vnd.viewcvs-markup
Given the current ROOT development plans, there is no chance to add
support the features proposed by CMS. On the other hand, René claimed
that learning to write custom streamers should take a short amount of
time.
Though all agreed that schema evolution would feasible with the approach
proposed by René, yet concerns were raised:
* need to train a large number of data formats
developers about writing custom streamers with
large load on training on the offline group
* introduction of explicit dependency on ROOT
in CMS data formats library
* possible error-prone code with explicit management
in the C++ of the different version cases and proper
ordering of data member in the file
CMS will reconsider the proposed approach, given the constraints
presented during the meeting. A possible alternative approach
will be writing a CMS-specific layer that uses the proposed ROOT
custom streamer internally shielding the details of ROOT streamers
to the data formats developers.