DDM Integration Testbed

At last Wednesday's meeting we agreed on a strategy for testing new releases of DQ2 0.3 on the US ATLAS computing facility, in the context of the Integration Program. This strategy includes:

1) Identification of a dedicated set of integration resources from the facilities project that would comprise a platform suitable for system-level functional and scalability tests of DQ2 updates. It would include:

• Tier1 servers for DQ2 site services and LRC
• One or more Tier2 sites for DQ2 site services and LRC
• A development Panda server at Tier1
• Other dedicated servers, eg. hosting an FTS service, as appropriate

In each case there is an associated backend storage resource which may or may not be a production SE, as appropriate.

What it does not necessarily include are DQ2 central services; it may attach to the production instances of these services, or development instances, as appropriate, in both cases provided by the CERN team.

2) This testbed would be independent of production resources that are in use for Panda, AOD replication, and serving datasets to physics groups.

3) A set of functional tests and benchmarks for a specific release would be established before recommending updates to the production instances of DQ2 site services and LRCs. These tests would include:

• DQ2 functional tests for subscription handling and

• submission of Panda jobs through the Panda development

4) Coordinated updates of the production US ATLAS DDM infrastructure as decided pending results of the integration process. The updates to the production infrastructure are recorded and tracked using processes established in the integration program.

5) Extensions to the integration testbed may occur to sites outside the US ATLAS facility, including possibly other Tier1 sites, as appropriate.

-- RobertGardner - 01 Aug 2007