Mark,

There are three (3) use cases (integration models) OCCI/CDMI can support. We have a few previous emails on this subject.

1) Block Device using a block device as a backing store. In this case the block device is a physical block device including iSCSI, FICON Disk, Serialized SCSIs or any other protocol supporting blck data interchange.
2) Block Device using a files device as a backing store. In this case the block device is a physical file including NFS, SAMBA, or streams based files systems.
3) A file system using a files device as a backing store. In this case the block device is a physical file including NFS, SAMBA, or streams based files systems.

These model requires the following data:
A) A URI indicating the source of the backing storage.
B) Optional credentials authorizing access to the backing storage.
C) A URI representing the integration point to the cloud compute resource/OS

OCCI also requires a unique ID to identify the instance of the data defined above.
D) A URI representing the instance of the storage resource

The last version of the OCCI Infrastructure specification limited storage to domain specific attributes; occi.storage.size and occi.storage.state. It inherits, the occi id from the OCCI core "kind"

CDMI can use the OCCI ID (see 'D' above) to identify the occi storage instance. CDMI data representing the storage source (see 'A' above) must be placed in a OCCI link (associated with the OCCI storage resource instance). Definition of how storage presents to the OS/cloud VM (see 'C' above) or any possible credentials (see 'B' above) has not been decided.

I hope this helps. These changes are very recent, and IMO,before we sink any more investment in the storage resource, we need to see how the new OCCI Core specification floats.

cheers,
gary

On 11/29/2010 7:50 AM, Mark Carlson wrote:

Is there a new version of the Infrastructure document that lists this?
Presumably this is the "compute" resource and the ID is a URL, right?

Also in the "storage" resource is it possible to add an optional "CDMI
Object ID" attribute?

-- mark

On 11/29/10 6:16 AM, Andy Edmonds wrote:
>From Mark and the CDMI guys.

In regards to Question 1 (below): In the context of the core specification it would be now more correct to say "the OCCI Entity sub-type's instance ID" as opposed to "OCCI virtual machine ID".

Andy

OCCI folks,

With a new OCCI version approaching finalization and CDMI about to go on to
ISO standardization, I thought we should revisit the OCCI/CDMI integration as
documented in CDMI and OCCI.

Question 1: Is "OCCI virtual machine ID" the correct term we should be using?
(If not, we can still "errata" CDMI to call out the correct term)

Question 2: Looking through the latest OCCI spec, I did not see a reference to
CDMI anywhere. Can we get a similar description as below pointing back to
CDMI to close this loop?

Thanks,

-- mark
_______________________________________________
occi-wg mailing list
occi-wg@ogf.org
http://www.ogf.org/mailman/listinfo/occi-wg
_______________________________________________
occi-wg mailing list
occi-wg@ogf.org
http://www.ogf.org/mailman/listinfo/occi-wg