Meeting Object:

WP4 workshop

 

 

Author:

Olof Bärring

Partner:

CERN

 

 

Meeting Date:

27 – 28 November 2001

Meeting Place:

CERN

 

 

Attendees:

Gaetano Maron (INFN), Andrea Chierici (INFN), Marco Serra (INFN), Enrico Ferro (INFN), Massimo Biasotto (INFN), Michele Michelotto (INFN), Thomas Röblitz (ZIB), Lord Hess (KIP), Paul Anderson (PPARC), Alexander Holt (PPARC), David Groep (NIKHEF), Kors Bos (NIKHEF), Lionel Cons (CERN), German Cancio (CERN), Jan van Eldik (CERN), Jan Iven (CERN), Olof Bärring (CERN), Julian Blake (CERN), Thorsten Kleinwort (CERN), Sylvain Chapeland (CERN), Bob Jones (CERN)

 

 

 

1. Decisions

1.1. Handling of input/output sandbox

The Resource Management task should coordinate the WP4 replacement for the handling of the input/output sandboxes with WP1.

1.2. Architecture issues

The architecture presented in the DeliverableD4.2 was accepted as a working model for the developments up to PM22, with the following addendums:

2. Information

2.1. WORKSHOP opening and final agenda (Olof)

Olof proposed the agenda for the workshop. He said that it has been decided to only have two DataGrid project meetings next year:

Olof said that he thought there was less need for special WP4 workshops outside the DataGrid project meetings next year. It was decided that no special WP4 workshops will be scheduled for 2002 and that this decision will be revisited at the WP4 meeting during the DataGrid project meeting in Paris, beginning of March 2002. On the other hand, tasks with distributed effort (e.g. the Installation Management task) are expected to organise their own workshops as appropriate.

The daily coordination of the release 2 developments will be done through the bi-weekly WP4 phone conferences, where each partner and task leader are expected to participate. Olof said that he wish to have a progress report from each task at those meetings and a special agenda item will be added for that purpose.

2.2. Integration reportS

2.2.1. CERN (Olof)

Olof gave a presentation on the integration of WP4 middleware, Globus and DataGrid middleware in general at the CERN testbed. Unfortunately Olof lacked information from October so some details may be missing from the report.

2.2.2. INFN Testbed report (Enrico)

Enrico gave a presentation on WP4 middleware integration at the INFN testbeds.

2.3. REports on Globus overlaps

2.3.1. Gridification task (David)

David gave a presentation on the Globus overlaps with the Gridification components. A written report is also available.

2.3.2. Resource Management task (Thomas)

Thomas gave a presentation on the Globus overlaps with the Resource Management components. A written report is also available. In the discussion that followed, one open issue emerged: if the Gridification task uses a modified version Globus gatekeeper it would be good if the Request Handler component could implement the Globus job manager interface. It was left to the Gridification and Resource Management tasks to resolve that issue internally. Another issue that arose was the management of the input/output sandboxes: in the current release this is handled by WP1 software that requires outbound IP access to all CPU nodes. It was decide that the Resource Management task should contact WP1 to coordinate the WP4 replacement of the current interim solution.

2.4. Architecture issues (German)

German listed the open issues on the WP4 architecture presented in the Deliverable D4.2. He had received comments mostly from Paul and Thomas. German also reported that the announced brainstorming on some of the architecture issues had taken place the day before the workshop. The conclusions from the brainstorming were presented to the workshop. The workshop then decided on a few addendums to D4.2, see section [1].

2.5. Global Specification (German)

The workshop went through the Global specification document circulated in mid-November. A number of modifications were considered necessary, where the most important were:

The project technical coordinator, Bob Jones, said that it would be good if WP4 could specify exactly what will and will not work for Solaris. It was thought to be too early to add such precise information to the global specifications.

2.6. Report on resources (Olof)

Olof presented the input he had received from the WP4 partners together with a proposal for how those resources should be allocated to the different tasks. Gaetano pointed out that the INFN efforts quoted was still subject to modifications because it depends on if a new person can be hired. However, Gaetano confirmed that the INFN effort is likely to fall between 2 and 2.5 FTE.

The efforts table originally presented by Olof contained several errors and columns for “integration” “architecture” and “administration” were missing.

2.7. Subsystem development plans

2.7.1. Installation management (German)

German presented the draft development plan for the installation management task. The plan had been elaborated after a brainstorming with members from CERN, INFN and Edinburgh. A long discussion followed, where the baseline was:

2.7.2. Configuration Management (Lionel)

Lionel had no slides, so here follows an attempt to catch what was explained on the white board:

 

 

 

 

 

 

 

 

 

 

 

 

 


Figure 1: (approximate) figure drawn by Lionel.

Abbreviations used in Figure 1:

 

Lionel presented the most important items in the development plan:

The developments of the CLI and GUI are outside the scope for PM22.

Priority ordering:

Estimated resources:

2.7.3. Monitoring (Olof)

Olof presented the development plan for the monitoring task. In the discussion that followed it was pointed out that a prototype of the Monitoring Sensor Agent (MSA) would be needed earlier than foreseen. The following step should probably be added to the plan:

2.7.4. Gridification (David)

David presented the development plan for the Gridification task. He pointed out that the initially available efforts will be low (~0.5FTE) but it is expected to rise to about 1FTE by summer. David wondered which other WP4 tasks had planned to use the FLIDS component for in their PM22 developments.

2.7.5. Resource Management (Thomas)

Thomas presented the development plan for the resource management task. The plan appeared reasonable to the workshop.

2.7.6. Fault Tolerance (Lord)

Lord presented the development plan for the fault tolerance task. Lord said that a lot of developments are foreseen for the short term because of a substantial additional effort from students is available up to end of January 2002. Thereafter the available efforts will decrease to about 1FTE for some time.

3. AoB

None