Minutes of the Data Design team telcon, 23rd November 2004

0) Summary of Actions
- Mark to look at next version of GFS-WG document.
- Mark to start revising draft document for UVa implementation.
- Manuel to see if Mark can get ahold of draft implementations of GFS/VFDS. 

- Dave to propose change to time slot

- Focus on metadata interfaces for services and map them onto scenarios
- Push on Architecture work.


1) Early discussion

Roll call:
Mark Morgan, UVa
Allen Luniewski, IBM
Ted Anderson, IBM
Manuel Pereira, IBM
Dave Berry, NeSC (chair and note taker)
Fred Maciel, Hitachi
Susan Malaika, IBM
Simon Laws, IBM
Leo Luan, IBM
Shannon Hastings, OSU     

The minutes of the 16th Nov meeting were approved.  Thanks to Mark for recording those minutes.

The agenda was unchanged.

2) Update on outstanding actions
- Mark to look at next version of GFS-WG document.  (ongoing action: the document was only released this morning)
- Mark to start revising draft document for UVa implementation.  (Ongoing action).
- Manuel to see if Mark can get ahold of draft implementations of GFS/VFDS.  (Ongoing: need to go through process at IBM.  Probably a binary version).

- People review scenarios for metadata to find errors and mistakes.
- Simon to describe general pattern and look at metadata people have suggested.  (Some progress - see below).
- Dave to push on Architecture work if he can.  (No progress).

3)  Update on Metadata scenarios (15 min)

Simon: new version of document now on GridForge.
Some interesting points arising.
Management not covered (no-one assigned).

Location of metadata: Metadata is often stored in two places.  Metadata particular to a service is located with the service; metadata in the registry often concerns additional information or relation between registry and service.

Shannon et al. listed possible questions regarding metadata.

Simon has listed types of metadata, but hasn't categorised it yet.  Dave Pearson produced a categorisation earlier, separting metadata into (1) data about data and (2) data about relationships.

We have two levels of interface to manipulating metadata:
 - simple: create, read, update, delete
 - advanced: publish, classify, discover, subscribe, manage

Next steps: focus on interfaces for services and map them onto scenarios.  

Mario's has raised question about semantics.  Shannon notes that the word is overloaded.  TO what extent can you automate the interpretation of semantic information?  Probably moving into the Semantic Web space.  We want our layer to be the base layer that the semantic Grid people will use.

Dave noted that the EGEE architecture document says that some data storage services do not hold the metadata about the data they store.  Hence EGEE centralises metadata in the registry rather than the data services.

Link to wider architecture: what kinds of metadata do we expect to encounter in other parts of the architecture?  E.g. if you have one piece of data derived from another, does the metadata also have to be updated?  We need be clear about the functionality that we aren't going to define.


4) BasicFiles and Contexts

No progress to report, but the new version of the GFS-WG is now available.

GFS were interested in the relationship with the naming design team and OASIS activity.  We suggested they join the naming design team calls.  Mark will forward details of call.


5) Plans for strawman architecture

Need to find time to do the work!  Possible sources: DAIS, GFS & GSM, EGEE, our metadata studies, our existing input materials.

Dave suggested a number of conceptual layers: data movement, replication, and federation.  This needs to be developed further off-line.

Fred raised the question of What we consider to be a resource?  As you go deeper into the architecture you bump into the question of what is a resource that you are manipulating?  E.g. in CMM, it wasn't clear whether data (as opposed to a data service) is a resource.

In OGSA v1, a resource is defined to be a manageable entity.  This raises the question, What does manageable mean?  It means you can introspect on it (e.g. WS-Resource).  There are some open questions that will be addressed in OGSA v2, e.g. is Management enough?  What about deployability, etc.?   Resource Management = e.g. WSRF + WSDM + operations and methods.

There is also a higher-level of management, that doesn't manipulate the resources themselves.  See CMM documents (just gone through 30-day review).

 
6) Wrap up

DONM: GFSG also meets at this time slot.  We propose to revert to previous slot (in UK and USA time).