Re: RE: GLIF subgroup on Global Lightpath IDs

[Ronald van der Pol <Ronald.vanderPol@xxxxxxxx>]

On Wed, May 28, 2008 at 19:11:47 -0400, Tom Lehman wrote:

> Ronald and All,
> 
> I discussed some of these concepts with some working on the perfsonar
> distributed lookup service.  I think there are still some concerns and
> unresolved issues with the use of totally random global ids.  Similar to
> what we were discussing with respect to distributed information services. If
> the global ids are totally random, it is not clear how you will ever know
> where in the distributed lookup service to go to get more information.  The
> result may be that we will require the distributed lookup up service
> distribute the ids globally.  

Actually, I do not have a strong preference. I have some concerns about
the total size, but thinking about it this is only a minor point. We
*could* use the global identifier in the path trace bytes. That *may*
have some operational benefits. I am not sure about the path trace
limits. It's either 16 or 64 bytes. The OME6500 supports 64 bytes,
so that would not be an issue. The internet2 based naming scheme
would work for 64 bytes. It won't for 16 bytes.

Thomas, Lars, which naming scheme do you like best? I really want to
reach consensus soon. We see the operational need for global identifiers
daily.

> That is the main reason we decided to leverage the inherent hierarchy in the
> dns name system, so that there is an established way to present an id to any
> point in the distributed lookup service, and it is possible to find your way
> back to the authorative information source, without having to distribute
> data globally.

Having a meaningful part in the global identifier is probably a good
thing. On the other hand, if it is just a key to a local name with
a meaningful syntax, it won't matter that much.

> I suggest that we have some more discussions with the perfsonar/distributed
> lookup service community before any decisions are made, and make sure we
> have do have a global id scheme that will play nicely with the distributed
> information services. 

That sounds like a good idea. Can you start a thread with the proper Cc's?

> Perhaps in this document we could lay out the pros/cons of random number
> based ids and ids which leverage the dns naming or some other topology
> schema hierarchy, and solicit input from the perfsonar subgroup, that was
> also established at the last GLIF meeting?

Thomas and I are in that group. But we are only working on the E2E
monitoring part. We have not discussed naming services yet.

	rvdp

Attachment: signature.asc
Description: Digital signature