DRAFT Minutes of the LCG SC2 meeting,
25/11/05
Present: Jean-Jacques Blaising, German
Cancio (secretary), Tony Doyle (via VRVS), Matthias Kasemann (chair), Marcel
Kunze, Eric Lan�on, Gerhard Raven
(via VRVS), Les Robertson
Apologies: Wisla Carena, Jim Shank, Albert de Roeck
Minutes of the LCG SC2 meeting, 25/11/05. 1
Organisational
matters. 1
News from the MB.. 1
LCG Status Report Review (Q3 2005) 2
Middleware
Area. 2
Applications
Area. 3
Fabric Area. 3
Grid
Deployment Area. 4
General
comments: 4
AOB.. 5
- The previous minutes (link)
were approved.
- Next SC2 meeting (Fabric Area
focus meeting): Since the LCG project, including the Fabric Area, has been
recently subject to a comprehensive review by the LHCC, it is decided to
cancel the next SC2 meeting scheduled for Friday December 16.
News from the MB
- Les reports from the Management Board
held on November 22.
- The
feedback from the Comprehensive Review was discussed. The subjects
included (see here for details): Lack
of quantitative data for LCG; CASTOR2 issues; 3D project status.
- VO boxes: A meeting with the
experiments will be organized in the second half of January, in order to
clarify issues related to the VO boxes.
- In response to a question by Matthias,
Les confirms that the evolutionary approach, which was chosen for
migrating from LCG-2 to gLite, is considered a success. All required EGEE
components are now part of the LCG software distribution, with the
exception of the Workload Management System that will be integrated in the
near future. The LCG software distribution will be renamed to gLite, and
will contain components from different sources like EGEE, VDT and LCG.
- The new ETICS project led by Alberto di Meglio is aiming for a
standard packaging and distribution system. Its relationship with existing
solutions like Yaim needs to be further investigated.
- Service Challenges: The failed SC3
throughput phase will be repeated in January. For this exercise, a new
version of dCache will be used. The service phase is progressing
reasonably well.
- CASTOR2: A Readiness Review is planned
for March 2006. No major problems have appeared for the last 3-4 weeks.
Matthias asks after the status of the �small file� problem. The current
approach followed by the experiments is to concatenate small files prior
to writing them to CASTOR.
- 3D project: A plan exists for deploying
3D. While services will be available already in March, the production
status will be reached by September. The setup will vary by experiment:
CMS is looking at FroNtier and Squid, whereas ATLAS/LHCb will use Oracle Streams.
Tier-1 sites might have to support both solutions.
- It has been recommended to disband ARDA
and to replace it by a more general support area for experiments.
- The relationship between EGEE and OSG
has improved significantly. There is not only a good attendance by OSG
people in the LCG MB but also efficient technical contacts between both
projects. Via its participating sites, OSG will report indirectly to LCG.
Also, the NorduGrid Facility will interface to LCG using gateways, which is
being discussed in the GGF. NorduGrid participated in the SC3 throughput
phase with good success.
- The MB Working Group, which addressed
reporting, monitoring and internal reviewing in LCG, has concluded its
work.
- A tabular, more detailed reporting format
will now be setup.
- Milestones for regional centers will
be included. Experiments will report through their respective Task
Forces.
- The reporting will be monthly; a
summary will be produced on a quarterly basis.
- There will be a series of ad-hoc internal
reviews with external participation; the first review will target CASTOR
in March.
- The Applications Area will continue
with regular internal reviews.
- There will be 1-4 annual service
readiness reviews; the first one take place prior to the startup of SC4.
- The first MoU
is likely to be signed by Academia Sinica (Taiwan) on
December the 6th.
- Tony
reports about his e-mail interaction with Frederic and Massimo.
- AliEn:
Tony mentions that AliEn has been adopted by the PANDA experiment (GSI).
Matthias emphasizes that there is no maintenance or support agreement with
CERN or LCG.
- ARDA:
- With
respect to the move of ARDA activities towards the umbrella of
experiments, Les explains that ARDA is funded 50% by EGEE and 50% by
CERN. On one hand, the experiments request that the work done by this
team should not be restricted to distributed analysis. On the other hand,
a tighter integration with the work of the EIS team is desired. Thanks to
a contribution by INFN, the size of the EIS team will be increased from
four to eight FTE�s.
- In
LCG phase I, two ARDA people were assigned per experiment. This model has
been subject to critics, because in some experiments, it took a long time
to converge on useful work. In order to improve efficiency, a new model
has been agreed for phase II. A total of 12 FTE�s (6 funded by CERN and 6
funded by EGEE) will be working on common, cross-experiment solutions.
Examples of such solutions are GANGA,
PROOF or �VO box� services. Massimo is now talking to the experiments;
convergence needs to be reached prior to the deadline for the EGEE-2
Technical Annex.
- Progress
has been achieved on the Pre-Production Service.
- Tony
considers the Distributed Analysis area being of the greatest concern. As
an example, Tony mentions the User Interfaces. The current GUI�s are very
large, in particular when comparing them with the light-weight and modular
packaging of NorduGrid equivalents.
- Les
informs that Claudio Grandi (INFN) has been
appointed as the new Middleware Area manager succeeding Frederic Hemmer.
- ROOT-SEAL merger: Gerhard highlights
that the integration is going in the right direction. A support agreement
has been reached; only a few SEAL libraries need to be maintained
independently from ROOT.
- Manpower in SPI: The manpower reduction
results from the move of manpower exclusively dedicated to SPI to more
shared roles. Gerhard considers it appropriate to have developers
participating part-time to common infrastructure activities, since this
encourages developers to use common tools.
- Jean-Jacques highlights that at the
last AA meeting, an agreement has been reached to provide a
ROOT-independent Linear Algebra module, which consists of the MathCore and the �small matrices� package, by December.
�
Missing milestones: Two milestones listed in the
Q2 report (1.2.3.6 � �Sub-station commissioned� and 1.2.1.4.1 � �Definition of
T0 building blocks�) are not found in the Q3 report. Bernd clarified by e-mail
that 1.2.3.6 was fulfilled on September 30, and that the date assigned for
1.2.1.4.1 in the Q2 report was wrong: its real due date was November 30 (link).
�
Proposed new milestones: A milestone for the
interface between CASTOR and the DAQ needs to be added, which is already taken
into account by the new milestone draft. With regard to the requested milestone
for the ALICE DCs, Bernd considers this being a
level-3 milestone, which is thus not shown in the Quarterly Report.
�
CASTOR-2 is of concern, because it gives the
impression of not yet being fully production quality. Marcel and Wisla
suggested to break up the CASTOR milestones into smaller steps. This should
include not only milestones on functionality but also on testing and stability
in order to harden the product. Les clarifies that the goal is to get CASTOR-2
ready for production by involving the experiments and this will include the
identification of contact persons in the experiments, the agreement of
schedules and the definition of corresponding milestones. No further problems
have been discovered since more than one month, but the experiments need to
make a heavier usage of CASTOR-2. In particular, the experiment production
activities still need to be migrated from CASTOR-1 to CASTOR2. The
problem-fixing rate is considered as very good, but there are still performance
issues to be looked at. The 750MB/s challenge will be completed before
Christmas and will use newly arrived IBM equipment that is being installed and
configured. As a consequence of an IT-internal reorganization, the CASTOR and
ELFms development teams have been joined, and vacancies for CASTOR development
are being prepared.
�
Eric points out that there is a cost increase of
up to 20% because of the augmented memory requirements for experiment jobs. The
maximum size has gone up from 1GB (as planned for in the TDR) to 2 GB. Les
informs that there are ongoing discussions on this subject in the GDB.
�
Eric expresses his worries about the SC3 set-up
milestone (1.4.7.2), because the full throughput goals were reportedly not met.
He wonders whether a new milestone needs to be added for this.
�
Reporting from OSG: Currently there is no
reporting channel, but Ian clarified that this has been foreseen for phase II.
�
OSG-EGEE interoperability: The submission from
EGEE to OSG has been tested, but the required reverse submission from OSG to
EGEE is still outstanding. Les explains that the schedule for interoperability
milestones has been discussed during a joint OSG-EGEE workshop in September.
�
3D service: The plans for the 3D service (see
also News
from MB section) have been discussed at the last
GDB meeting. The experiments still need to define their plans on how and when
to use 3D; discussions for SC4 planning will take place at the Mumbai meeting
in February. One site has expressed objections about running both, FroNtier/Squid and Oracle Streams.
- The renaming of the LCG middleware
distribution into �gLite� has created some confusion, in particular with
regard to modules, which are part of the original gLite package but not
part of LCG-2 (like FiReMan). It is felt that this name change has not
been announced widely enough. Les informs that this decision was agreed at
the EGEE Management Board meeting in Pisa
and will be further discussed at the GDB meeting in January.
- Points for the POB:
- Applications Area: Most concerns from
the last review have been addressed. The FLUKA situation has not yet been
completely resolved.
- Fabric Area: CASTOR remains a concern,
even if good progress has been achieved.
- The new project reporting structure
will help to achieve a closer involvement of remote sites.
- The work of the LCG SC2
committee ends with this meeting. Matthias expresses his thanks to the
present and past SC2 members for their hard work and dedication.