Minutes of the OGSA-D WG telcon, June 7th 2006


0) Actions arising

* New actions

- Stephen to ask Luc's provenance team why they have chosen a separate provenance store.
- Crhis to continue work on GFS scenario

* Ongoing actions

- Dave to contact OGSA WG about the dates in the roadmap document.
- Stephen to carry on chasing Luc about the provenance scenario.

- Stephen & Dave to review scenarios doc from a storage perspective.
- Dave to draft metadata catalog scenario & services (cf. SRB)
- Ann to check the Access section for suitability for file replication.
- Section authors review security aspects of their sections.


1) Early discussion

* Roll call

Dave Berry, NeSC (note taker)
Mario Antonioletti, EPCC
Stephen Davey, NeSC
Allen Luniewski, IBM
Chris Jordan, SDSC

The minutes were approved.


2) Action report

- Chris to draft GFS scenario [Done]

Other actions ongoing (as above)


3) Review of GFS scenario

Chris noted that some of the text from the scenario section belongs in the functional design section.

The GFS could sit on top of the replication service, with the GFS controlling when transfers are required.

Step 4 of the scenario needs to be clarified to explain that the RNS virtual directories contain the "primary" copy of the data.  It does not create one virtual directory for each copy of the data.

Step 6 creates a second RNS directory for "exported" files.  This does not preclude the existence of replicated versions in an underlying replication system.  The second RNS directory is primarily created to provide a separate access control mechanism.

We discussed whether caching services would be useful additions to the implementation instead of replication/transfer, or could even use caching instead of replication.  Step 5 seems to be a good place to use caching.  This scenario might allow caching strategies based on predictive uses, modelling typical access patterns.

The scenario needs diagrams and also an explanations of which interfaces are used at each step.

There may need to be some kind of management interface.  Extra text could be added to clearly motivate the need for such an interface.


4) Review of Provenance scenario

We discussed whether provenance data should really be stored in a separate store, rathe than with the data.  We wondered whether the use of a separate store was an arteface of existing (file-oriented) systems and whether we have the opportunity to encourage the approach of storing provenance data as metadata of the data object itself.

The system would have to handle updates, augmenting the metadata when changed.  It also has to make the provenance material to be auditable.  Also, all data would have to embed the provenance information.

Stephen to ask Luc's provenance team why they have chosen this path.


5) Wrap up

- June 14: Review scenarios doc (Apologies from Allen).
- June 21: Short call (half an hour later). Review arch doc?
- June 28: Provenance? 
- July 5:  No call?
- July 12: ?