Minutes of the OGSA Data Design Team telcon on 19th January 2005.

0) Actions arising
Simon to mail the list about different types of metadata (data resource vs. service).
Dave to check with Dave Snelling when Basic Profile is ready for review.
Discuss time slot in e-mail.

1) Early discussion

* Roll call:
Susan Malaika, IBM
Simon Laws, IBM
Andrew Grimshaw, UVa
Mario Antonioletti, EPCC
Dave Berry, NeSC (chair and note taker)
Allen Luniewski, IBM

No minutes were taken for the meeting of the 12th, because it only attracted only 3 participants.
        

2) Update on outstanding actions
- Simon to flesh out his question about different types of metadata (data resource vs. service) for discussion in a broader group.  
Simon has written this and will send to list later.


3) Update on OGSA Profile work
EMS are focussing on establishing standards from which a candidate profiles can be created.

We discussed the role of profiles: are they to specify interoperability between implementations of existing standards or to specify interoperability between standards?

We suggested an initial data profile on basic stuff:  RNS, BasicFiles, DAIS.  First two address scenario of accessing remote files (or files that have been mapped in).  RNS + DAIS for issuing queries.  Possibility of reading results of DAIS queries using BasicFiles interface.

More advanced profiles would include transactions, replication, metadata, service connections, federation, derived types from BasicFiles (e.g. XML, next element).

DAIS would work well with RNS.  RNS would also be useful for dynamically-constructed federation services.

We need to consider how file transfer services & protocols.


4) Update on upcoming meetings 
- UK meetings.  This will be on the afternoon of Feb 2, at Imperial College London.
- OGSA F2F, Chicago, Feb 14-16.
No-one was planning to attend, although Susan could call in.  DAIS/DT will review Basic Profile prior to meeting.  Dave to check with Dave Snelling when Basic Profile is ready for review.


5) Architecture issues
- Is data transfer a service?
Postponed until next time.

- Is there any standards activity w.r.t. replication?  Should we organise a workshop / F2F meeting on this topic?
Many reasons for replication: performance (better done by caching), availability, data warehousing (abstracting subset of data out of set of databases into the data warehouse). Replication is a use case for Info-D.

- Do we need a standalone cache service?
A distinct cache service is very powerful for performance gains, with option to select cache coherence options.  Can use replication but caches give more options for maintaining coherence.

Cache services would not be visible to the application.  Caches and owned by either the resource or by the user/application.  Authorisation will allow owners to control who can manage the service, alter the coherency, etc.


6) Wrap up (5 min)
      DONM: Next week.  There was a request to change the time slot, which we will discuss in in e-mail.