LCG Management Board

 

Date/Time:

Tuesday 8 November 2005 from 16:00 to 18:00

Agenda:

http://agenda.cern.ch/fullAgenda.php?ida=a057064

Members:

http://lcg.web.cern.ch/LCG/Boards/MB/mb-members.html

 

(Version 2 - 14.11.2005)

Participants:

A.Aimar (notes), D.Barberis, L.Bauerdick, I.Bird, K.Bos, F.Hernandez, N.Brook, T.Cass, Ph.Charpentier, Di Quing, I.Fisk, J.Gordon, B.Gibbard, M.Lamanna, E.Laure, H.Marten, P.Mato, G.Merino, B.Panzer, J.Shiers, L.Robertson (chair)

Next Meeting:

Tuesday 15 November 2005 at 17:00 will be a phone meeting

Approval of minutes (minutes)

Minutes approved.

 

Matters arising

Site plans

Received almost all site plans at the GDB in Bologna and converted them to the templates discussed last week. Converted them to the milestones templates and will be presented to the GDB asking sites to complete them (see wiki page). SARA, PIC and FZK had a first update of the plans. But all need to be completed.

 

Action: Tier-1 Sites (at GDB) should complete their plans following the guidelines of the presentation given at the GDB. (GDB presentation).

 

New version of the reporting proposal

Distributed the new version to the MB and there is one open issue about reporting of the Task Forces for the grid environments.

 

New proposal of the Task Force description (new proposal)
From the proposal:

While the higher level issues apply to all LCG services, independent of the grid on which they are provided (e.g. EGEE, OSG), lower level issues are covered by the Task Forces only for the EGEE grid, where there is a close and direct relationship with LCG at the operational level.

 

TF will provide the QR for all high level issues concerning the experiments’ use of the LCG services and also all other issues concerning EGEE. 

 

There was the open issue of how OSG and Nordugrid grid environments will report their milestones and progress, and on how they will receive requests from the Task Forces.

 

The proposal that the grid organizations be represented by the American centers supporting them (FNAL for OSG/CMS, BNL for OSG/ATLAS) was not accepted by ATLAS that would like the reporting line to the MB to be via the experiment.

 

CMS agreed that OSG reporting will go via I.Fisk who represents FNAL/USCMS software and computing project in the MB.

 

ATLAS will discuss how the reporting should be handled for resources outside EGEE and make a proposal to the MB.

 

The situation with NDGF (the distributed Tier-1 serving ATLAS and ALICE, and with Tier-2 facilities for CMS) was not discussed . These were no representatives of ALICE and NDGF present at the MB.

 

Action: ATLAS will send a proposal about reporting on issues concerning non-EGEE resources to the MB.

 

Action: CMS will send a similar statement to formally confirm the CMS opinion on OSG representation.

 

The experiments will report separately concerning their use of Applications Area services.

 

Announcements

No announcements

 

No meetings to report on.

 

Future of the Phase 2 Planning group

The P2P started on March of last year, to collect overall requirements for Tier-1 and Tier-2 centers, to provide input to the MoU taskforce.

 

The last meeting was in September to prepare the data that is in the Annex of the MoU.

 

The proposal is to stop the P2P group and to transfer the responsibility for overall resource planning to the MB. This will be done twice per year in early March and early September, prior to the C-RRB meetings.

 

The MB agrees to stop the P2P.

 

Chris Eck will continue to be the person to receive the regional centre resource information and update the information on the web. The MB will check it well before each RRB.

 

Action: Les Robertson will pass the decision to the P2P group.

 

Storage Working Group

It was initiated during the Phase 2 Planning meetings and became a joint LCG and  Hepix working group. The report was scheduled to be given at the HEPiX workshop meeting, but was postponed because the report was not finished. As the P2P group will now stop the proposal is that this activity will be handled by the GDB from now on.

 

The report of the working group is still being debated, with discussions going on inside the working group. A presentation is scheduled in the GDB meeting on 8 November (see presentation) and will be discussed there (see GDB minutes).

 

Is important because many sites are purchasing lot of disk and tape drives capacity in this period. There is also a proposal for a full day storage workshop, also joint with HEPiX, connected to the GDB meeting in April.

FZK felt that a study was needed and that is why the SWG was started. Now it is not urgently required as FZK has started in parallel to purchase equipment.

 

 

Milestones for Phase 2 (see High Level Milestones)

The milestones for Phase 2 were distributed to the GDB in October and to the MB last week.

 

No feedback from GDB.

 

From MB members:

  • There are no milestones for Tier-1-Tier2 activities and Tier-1-Tier-1 activities.

 

Some emails have been circulated but a more formal proposal is needed.

 

F.Hernandez for IN2P3:

A project covering inter-working with Fermilab has been approved. IN2P3 will provide one or two milestones based on this work once it has been better defined. The transatlantic link Lyon-Fermilab will become 10 Gb in 2007 (currently 1 Gb). The project partners still have not met. Also a link Japan-Lyon is being planned. Some dates will be proposed in the following weeks.

 

J.Gordon for RAL:

RAL also has Tier-1-Tier-2 testing going on for which milestones can be provided.

 

K.Bos for SARA/NIKHEF:
Is important to have milestones defining when Tier-1 centres will require to read from tape”in addition to data recording. The service challenges have so far only covered writing data to tape. A milestone should be set for the verification of mixed use of the tape service.

 

D.Barberis for ATLAS:

For now experiments have not defined explicitly when this will be needed, but for sure it is an issue and will be tried during SC4, beginning of next year.

 

L.Robertson:

Sites should plan tests to check their capability for the full Tier-1 services, including simultaneous tape read and write access. These tests need to be planned before the experiment begin to do their tests.

 

Action: Tier-1 Sites to set milestones on their known Tier-1-Tier-1 and Tier-1-Tier-2 testing activities. 

 

J.Gordon for RAL:

RAL did some tests last year, but they are changing to use new robots now, and therefore the tests will need to be redone. One could also try destructive testing: trying to push it until a bottleneck is reached in terms of performance and functionality.

 

L.Bauerdick for CMS:

CMS will need to write to tape and run analysis jobs at the same time.

 

G. Merino for PIC:

It is a key factor to have milestones based on real use cases, such as the read/write cases. Sites lack the real usage model from the experiments. Testing in SC3 is not yet very realistic. SC4 should be more realistic and have appropriate milestones.

 

L.Bauerdick for CMS:

CMS has been trying to simulate the Tier-1 scenarios. The tests are not the same as in the real situation but we need to start somewhere. The tests will be done with all Tier-1s for CMS in Phase 2 of SC3. For reprocessing there are performance goals for these sites, from disk to tape and vice versa.

 

L.Robertson:

We need to have these goals set as milestones with targets set by the experiments from their computing models, with the sites involved matching these requirements. He proposed milestones for each experiment, each milestone involving at least three sites.

 

D.Barberis for ATLAS:

Data rates and data flows are in the TDR.


K.Bos for SARA/NIKHEF:
In the Netherlands storage will be in the SARA Computing Centre and computing will be at Nikhef. How will the experiment usage match this architecture? More information on exactly how the work is performed in the Tier-1s would be useful, including detailed sequence diagrams”.


N.Brook for LHCB:
Experiments can give their needs, but cannot define the sites architecture – this is the responsibility of the site experts. The needs are all in the computing TDR.

 

L.Bauerdick for CMS:
This is a common problem and is important; in order to define the site structure one needs to know how they are used. Maybe a baseline could be defined, how many servers with high-throughput connections? And to how many CPU? In which context this should be done? CMS in the Tier-0 centre would like to understand how much data can be on tape and how much can be in file buffers or be deleted from buffers, etc. Early Spring could be a good time for these Tier-0 tests, and then do the same tests for Tier-1. Maybe a questionnaire would be useful.

 

F.Carminati for ALICE, via an email prior to the meeting:

Tier-0 and Tier-1 sites should concentrate primarily on recording in 2007. Reprocessing will become important only later.

 

D.Barberis for ATLAS and L.Bauerdick for CMS:

Reprocessing will be needed quite soon in order to check the data, and Tier-2s will want to start processing and reading from Tier-1s as soon as possible. So we need to be able to read back from day 1. ATLAS and CMS want to do some testing of this use case.

 

In general it seems that many things are not very clear to some sites. How do we extract a clear model from the information in the TDRs?

 

J. Gordon:

It is important also to know what is needed at the same time. How many of the use cases can run at the same time? We need to discuss more that just look at the TDR.

 

 

J.Shiers LCG Service Manager:

There will be some discussion of required transfer rates for SC4 at the GDB on 9 November.

 

There will be a Service Challenge Workshop in Mumbai in February immediately prior to CHEP’06.  The draft agenda is:

Day 1: Data management issues, SRM, dCache, Castor
Day 2: All other service to meet the MOU targets
Day 3: Experiments use cases for SC4.

This is an opportunity to enable the experiments and the Tier-1 centres to get up to speed on understanding the Tier-1 workload and data flows. Most of the Tier-1s will be present at the workshop (PIC is not sure, CNAF, NDGF, Triumf wer not present).

Another issue is to understand the Tier-1-Tier-1 traffic

 

J.Shiers, LCG Service Manager:
The primary use case is getting data out of CERN. For other transfers one needs to have clear use cases, for instance this is when you reprocess and redistribute the results to other sites. The first is a configuration issue for the channel, while the other is a new use case. The raw data does not move but a copy of the ESD/AOD needs to be distributed after reprocessing.

J.Gordon for RAL:
The light path could be switched at CERN to support traffic between Tier-1s.

D.Barberis for ATLAS:

Reprocessing can happen all the time even if nobody wants to do it more than once a year.

L. Robertson:

By February we need to have enough information on Tier-1 operations, and also Tier-1-Tier-1 and Tier-1-Tier-2 activities.

 

Action: Preparation for the Mumbai workshop – details should be agreed following the 9 November GDB.

 

Action: All MB members to provide more feedback on the High Level Milestones.


---------------
G.Merino was asking about the preproduction system in a separate email. I.Bird will reply outside the MB meeting.

 

 

Responsibility for accounting

At the last GDB the technical presentations on accounting were very useful and it was clear that there were sufficient tools and sites could manage to produce the data needed.

 

There now the need to take responsibility in really getting the accounting data from the sites: J.Gordon (RAL) accepted to do it.

 

It includes also decisions on which values are interesting, collecting data, normalizing the values, and distributing the accounting reports. Reports similar to those produced for the RRB will be sufficient for the MB, including both EGEE and non-EGEE grid sites.

 

A.O.B.

Summary of the Action List

 

All pending actions from the previous MB meeting have been executed.

 

Current actions:

 

Tier-1 Sites (at GDB) should complete their plans following the guidelines of the presentation given at the GDB. (GDB presentation).

 

 ATLAS will send a proposal about reporting on issues concerning non-EGEE resources to the MB.

 

CMS will send a similar statement to formally confirm the CMS opinion on OSG representation.

 

L.Robertson will announce the decision to stop the Phase 2 Planning Group..

 

Tier-1 Sites to set milestones on their known Tier-1-Tier-1 and Tier-1-Tier2 testing activities (IN2P3 and RAL stated specifically that they will do this).

 

All MB members - Preparation for the Mumbai workshop – details should be agreed following the 9 November GDB.

 

All MB members to provide more feedback on the High Level Milestone is needed.