LHC Computing Grid Project

Project Execution Board

Notes of the meeting of Tuesday October 14, 2003 – Final as agreed on November 4

 

Present:

Jean-Philippe Baud, Ian Bird, Marco Cattaneo, Chris Eck, David Foster, Frederic Hemmer, Bob Jones, Jürgen Knobloch, Alberto Masoni, Mirco Mazzucato (phone), Bernd Panzer, Gilbert Poulard, Les Robertson, Torre Wenaus (Phone)

Minutes of previous meeting(s) (September 23 & 30) and matters arising

The minutes of the last meeting were approved.

Summary of meeting on Replica Registration Service (Jean-Philippe Baud) ( transparencies )

Jean-Philippe summarized a meeting held at LBNL in September with representatives from JLAB, FNAL,USC, LBNL and CERN

with the main topic: interoperability between Replica Management systems. Tasks of replica management are selecting the Replica Catalog, getting SURLs of files to be replicated, allocating space (Storage Resource Manager), copying and registering files. Assessing the existing systems, it was found that there is no standard replica catalog and that the interfaces are different and that none of the authors was willing to change their interface or catalog.

As way out, an intermediate layer (Replica Registration Service - RRS) was proposed. A problem is that Globus has not implemented the notion of Grid Unique Identifiers (GUIDs) different from the logical filenames that is implemented in the EDG RLS. Access control would go via Grid certificate without distinction between read and write. In conclusion, the sites present at the meeting agreed to develop the RRS interface to their own Replica Catalog, although nobody has proposed yet to do the interface to the Globus RLS. It is proposed that GFAL and POOL use the RRS. This would simplify their code.

In the discussion, people were concerned about the extra layer and raised the question of scalability. It would have been preferable to change one or two of the interfaces such that they match exactly each other. Further discussions will take place next week. Les advocated a pragmatic solution without additional complication and to aim for a standardization process in the long term.

 

EGEE and LCG - Operations strategy (Ian Bird) ( transparencies )

The EGEE activity areas are services, middleware and networking (in the EU sense – meaning outreach). The areas are subdivided into 11 activities. The project management has been assigned: Fabrizio Gagliardi (Project Director), Bob Jones (Technical Director), Ian Bird (Operations Manager), Frederic Hemmer (MWare Manager), Malcolm Atkinson – NeSC (Dissemination & Training), Guy Wormser – CNRS (Applications Manager), Gabriel Zaquine – CS (QA Head), Fredrik Hedman – PDC (Security Head).

The operations structure is layered with one Operations Management Centre (CERN), five Operations Infrastructure Centres, about ten Regional Operations Centres and more than 50 Resource Centres. The initial service will be based on the LCG infrastructure and will also need a development service and a test-bed system. In the various federations (CERN, large countries and regions), a total of 14.3 M€ has been requested to fund the manpower. The project-wide compute power is expected to grow from 3000 nodes at the beginning to almost 9000 after 15 months. The available resources will support many virtual organizations respecting the specific allocation policies of the respective resource centres. Most centres share resources between different sciences while others (like CERN) are dedicated to a specific field.

Ian concludes that having a running LCG service is crucial to the start up of EGEE; EGEE should be operating the European grid infrastructure on behalf of LCG by end 2004 and that much work to is ahead to set up operations infrastructure and define the implementation plan – work needs to begin now.

 

RRB preparations (Chris Eck) ( transparencies )

Chris gave a preview of the main contents of the report on the LCG Project resources and prepared for a first discussion of LCG Phase 2 resources.

For Phase 1 (in the years 2002 – 2005)), about 220 FTE-years of personnel have been externally funded. In addition an average of about 70 FTEs (decreasing from now on if nothing is done) are provided by CERN/IT and EP. This is currently supplemented by about 10 people from the experiments. The Phase 1 materials expenditure has been reduced over the past 24 months from about 30 MCHF to about 20 MCHF. The funding is still short by 4.1 MCHF for the years 2004/5. Extra contributions are expected to reduce this shortfall.

Chris pointed out the difficulty of obtaining data from the regional centres on the planned capacity for the next 9 months. Currently, the requests by the experiments exceed the claimed resources as of April 2004.

For Phase 2 (Tier 0+1 installation and operation at CERN) there is currently a shortfall of 20.2 MCHF of which 4.4 are for materials the rest is for 105 FTE-years.

Report from the GDB

Alberto summarized the last GDB:

John Gordon has presented the status of Grid operations.

Ian Bird has presented the status of deployment of LCG-1.

Following a presentation by Ruth Pordes on GRID-3, the main discussion point was on grids interoperability. For the LCG applications will operate across federated grids. In Europe the EGEE project constitutes a grid federation, based on the same middleware as the LCG.  Specific effort on grid interoperability will continue to be necessary, working with the US grids and elsewhere to meet the specific needs of the LHC experiments.


Jürgen Knobloch