Present:
Dario Barberis,
Phone: Tony Doyle, Francois Etienne
Apologies: Nick Brook, Bernd Panzer, Dave Foster
The minutes of the meetings on the 4th and 11th November have been accepted without change.
Les and the application managers present gave their first impressions,
main points and action items after the internal and LHCC reviews of the LCG
project.
Torre based his presentation on the last slide of
his PPT-presentation.
David Stickland for CMS launched some
discussion on the decision-making process in the applications area, in
particular on the role of the architects forum (AF).
He felt that the planned move towards a design, implementation and management
as outlined in the LCG Blueprint should be made more visible in the planning of
the ROOT team. CMS considers such a move vital if we are to consider any real
change in the LCG/ROOT relationship as outlined above. CMS recognizes the
difficulty for ROOT to make such changes given the maturity of the product and
is thus concerned that the proposed change of relationship may be very
difficult to achieve.
CMS was surprised by the decision to develop a third build tool, APWORK,
rather than make use of either SCRAM or CMT. The process of making this
decision is quite unclear to CMS.
While CMS welcomes the work to enable LHC experiments to use FLUGG in
test-beam examples to give insight on the GEANT4/FLUKA differences/features,
they do not currently have an interest in a common geometry modeller such as
the VMC – they also made this clear in the RTAG 9/10. CMS does, however,
have an interest in an exchange geometry description, to which experiments and
software providers should develop appropriate interfaces to their internal
models.
David stressed that any work starting in the applications area must have
the commitment of a quorum of the LHC experiments with integration milestones
at an appropriate time-scale. He requests that the decision-making process in
the AA be strengthened with clearly defined decisions and explicit experiment
commitments.
Torre was surprised that the process of the AF
was not clear. He said that after a decision in the AF, preliminary notes were
distributed to the experiments and sufficient time for feedback was given. The
detailed notes and decisions of the AF are made publicly available from the AA
web-site.
The representatives of the other experiments in the
PEB felt that the current decision making process for the applications area was
appropriate and was working to their satisfaction. – It should of course
be made sure that the workplan and resources have to
stay within the scope as defined by the SC2. It was recognized that with time
priorities may change and that it was useful to go from time to time through
the projects first in the AA and then providing a summary for the SC2.
Les proposed that as standing agenda item, Torre
would report to the PEB major decisions taken in the AF. This proposal was
endorsed. The PEB agenda plan, tabled later in the meeting, foresees two PEB
meetings early next year to make a high level workplan
and resources review of the AA.
Also Ian used the last slide of his PPT-presentation.
Concerning organisational matters, Ian proposed to introduce technical
meetings of people doing the work also at the system manager level. Ian has
taken measures to appoint the requested LCG security officer – initially
ad interim. An important issue is to strengthen the collaboration between
regional centres.
On the topic of installation, it is planned to remove dependencies
between middleware components, to simplify the installation and to improve the
certification testing. In order to achieve this and to guarantee proper support
for another year, effort will be spent set up a proper build system and to
break dependencies. This effort will be launched soon after deployment of LCG-2
and should initially be decoupled from the EGEE effort in order to react
rapidly to the review.
Les has written up his personal summary and major conclusions in a document.
One point the experiments have to think about is how the requirements
for computing resources at CERN would be reviewed after the end of COCOTIME.
A major point was the question of collaboration of regional centres in
the context of the GDB.
Unexpectedly, a point of LCG support for MC generator development was
raised by a member of the LHCC (Michelangelo Mangano).
We should investigate and react in a rather short time.
The plan is to release LCG-2 by end of this week. The date has slipped
from mid-November because of newly discovered major bugs and another POOL/RLS
schema change. The goal is to deploy before the Christmas holidays expecting a
consolidated stable LCG-2 in mid-January. Some components will this time still
require re-installs. It was emphasized
by all that from now on in production mode backward compatibility was essential
and must be ensured. From now on no more schema changes affecting the users
will be accepted for LCG-2. Things may be different in the long term for major
new middleware (i.e. replace for LCG-2).
The storage access will start with SRM-enabled storage elements offering
packaged installations of Castor & dCache, where
Castor is ready now and the dCache version is
developed by RAL with the help of Fermilab. The latter
is expected to be ready within two weeks. Storage access will not be
distributed with LCG-2
but instructions and pointers will be provided.
GFAL has been tested at CERN and Fermilab and
will be in the distribution together with a user guide and release notes.
With LCG-2 the distribution of experiment software has been separated
from the middleware deployment. LCG now
provides the tools and mechanism for distributing experiment software but the
responsibility now lies with the experiment software managers.
David Stickland emphasized the importance to
have a stable environment where the experiments can install their software. Ian
announced that a set of machines running LCG-2 different from the certification
testbed will be available.
On a question of Federico on the support for MSS at regional centres,
Ian said that all sites having SRM will be supported though at the time being
there was no experience with the existing SRM interface to HPSS – tests
will need to be done in Lyon and Brookhaven. Anyhow sites interesting to
PEB
meeting schedule for the next few months (more
information)
Les provided
first ideas on the schedule of upcoming PEB meetings. Due to the lack of time
this will be discussed further by email.
Roles
& Mandate of the new PEB
Les went through a document he prepared on the composition and the roles
of the new PEB.
After agreement in the PEB the document will be sent to the POB.
Federico provided a number of comments, corrections and additions most
of which have been integrated into the document.
By the time these minutes are written, a new
version of the document has been sent to the POB with copy to the PEB.
Some discussion went on concerning the respective roles of GDB, PEB and
the experiments with respect to scheduling resources and planning of data
challenges as well as the choice of middleware.
It was agreed that the technical bodies such as the security group, the
architects forum etc. would make recommendations to be endorsed by GDB and PEB.
This time the quarterly report will be delivered to the SC2 by the end of January. The dates for the various steps in between will be discussed by email.