[Nsi-wg] describing connections between networks/ V.09

John Vollbrecht jrv at internet2.edu
Tue Jul 20 11:51:03 CDT 2010

Guy -

I reply quickly because I know we want to release this doc as soon as possible.  I apologize for raising this issue again, but I think it very  important for future work to try to show this concept clearly.  I hope my explanation is clear

TThe top picture in figures is 9 and 11 good. I do have a problem with the example networks in the lower picture of figures 9 and 11  because they show links between networks that do not belong to one network or the other.   

I do understand that in one case described by Kudoh San where a jumper exists between two nodes, and is not owned or managed by anyone, that such a connection might be possible.  However, the other two cases (link as a network by itself, link as part of one network) he shows do not have any resource between the two networks, and these cases are not shown in the examples - and for me these are the normal ways of connecting networks.

I try to explain the issue - .  my concern is how - really whether - to treat NML link resource in NSI.  In my mind  a link (in NML terms) needs  to be in one network or another because it is a resource and is scheduled by one NRM or another (the exception is Tomohiro's case 3 where it can be ignored).   NSI does not define links or nodes - as shown in the top diagram of both figures.   Showing links beween networks implies that the underlying implementation cannot be described without including links between networks, which is not what is shown in the top picture of each figure.

How do links and nodes fit the NSI network model?  The NML wg  deals with nodes and links and groups of nodes and links.  NML can describe the topology of links and nodes, and can create groups with  different characteristics.  An NSI (transport) network can be considered to be a NML group of resources that are available to an NSA.  It is a network object in NML terms.

What is the difference between a NSI network and a routing domain?  In my view the biggest difference  between an NSI transport network and a NML routing domain is that an NSI network edge might be the end of a link or the end of a node, while the NML routing domain includes interconnected nodes.   Connecting NSI Networks is accomplished by joining the edges of two networks to each other, while connecting NML routing domains is accomplished by inserting a link between a node in each domain.

The concept of an NSI Network does not include links and nodes.  An NSI (transport) network is a resource that can provide services under control of NSA.  One of the services is a connection service between STPs.  The capabilities of an NSI network are defined by a Transfer Function.    The NSI network is a network object in itself and can be connected to other NSI networks. [ Note that since a NSI network may internally be a set of nodes and links the  transfer function could be a topology of these nodes and links.]

We have chosen to call the edge of a network an STP.  A connection between NSI networks is created by joining STPs from each network. The name of this junction is not clear to me - I think we call it an SDP    NML group can take this set of concepts and describe them in NML - that is their job.  Our job is to describe the concepts.

How to picture this in your examples is a question,  You show ports on the edge of networks.  This is ok for me if ports exist on any resource (link or node)at the edge of a network.  This is the way the G.805 describes them.  It is not the way the NML currently describes them.   An alternative would be to call the edges by a different name - STP for example. 

Which terms to use is a hard call.  Port is a term that is well understood at physical level.   But port is a tough term for NML since for them it implies a node, and using it seems to imply that edges of NSI networks are all nodes.  I think the term STP at the edge and SDP between edges is the best I have heard so far - it presents the concept without confusing it with other established concepts.  You may be able to come up with a better way.  In my view we need to present the concept to NML, not try to map our concept into their terms.   Fitting this seems possible and  a NML task which they can only do when we present the concept to them.

I modify figure 11 to show these changes -- still using ports (because I think this is what you want to do) as the term, but implying link or node port.
Figure 9 has the same issues - and I sent an edited version of that earlier.  I could send again if you like.

regards -


On Jul 20, 2010, at 8:55 AM, Guy Roberts wrote:

> I have incorporated comments from Inder, John V, Chin, Kudoh-san into the new version 0.9:
> http://forge.gridforum.org/sf/go/doc15985?nav=1
> Guy
> _______________________________________________
> nsi-wg mailing list
> nsi-wg at ogf.org
> http://www.ogf.org/mailman/listinfo/nsi-wg

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/nsi-wg/attachments/20100720/756003df/attachment-0002.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: figure11.jrv.july2010.pptx
Type: application/vnd.openxmlformats-officedocument.presentationml.presentation
Size: 92599 bytes
Desc: not available
Url : http://www.ogf.org/pipermail/nsi-wg/attachments/20100720/756003df/attachment-0001.bin 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/nsi-wg/attachments/20100720/756003df/attachment-0003.html 

More information about the nsi-wg mailing list