[ogsa-d-wg] [acs-wg] Data Use Case Scenarios document

[Keisuke Fukui]

Hi Pete and folks,

pete at ziu.net wrote:
> 
> I read the current architecture draft ogsa-d-wg has up on their site, 
> and there appears to be no conflicts. They do mention non-master 
> replication in the context of "peer-to-peer". I refered to these a 
> little differently in terms of "push", "pull", and "bi-directional". 

That's excellent to find there is not conflicts between the specs.
Then, we can review what is potentially missing in our spec. Let's see
the diagram Sachiko will post to think about this.

> Mike, did you find the use cases up on ogsa-d or just through 
> communications with Steven? I am working to continue to add detail to 
> the replication presentation. I am attempting to define a 
> "ReplicaElement" structure which will have properties of a unique 
> instance id, replica id, the actual value of the data, modification and 
> author history, and security constraint data. This one structure should 
> be able to be used to represent a piece of data, a collection, a graph 
> (tree), and a database, for a collection of replicas. Hopefully I can 
> get this out before ogsa face2face to help stimulate the replication 
> topics.
> 

It's good to make sure our spec. doesn't conflict with the data architecture
replication. Pete, if you can drill down this a little bit more it will
surely be of help to us. You may be able to talk with someone from
ogsa-data team at the F2F.


Though I see the value of the data replication use case, I think that it is
best to leave the detailed normative description for the post version 1.
I'm still aiming for the first spec to be submitted at GGF15. To do so,
we need to have closure on the draft well ahead of Oct. I think we should
be focusing on making this deadline. Designing the ACS dedicated interface
for data replication will inevitably involve a lot of work, including review
of existing products. Considering that there are bunch of famous data
replication products in the market, including SRDF from EMC, and embedded
features in the RDBMs, we may not be able to design efficient interface for
it. We have plenty of time after version 1 is out to do this.

Let's discuss at the call about what we can do for our first draft.

-Keisuke