Joe,

Thank you for the great summarization on the technical approaches.

NVO3 applications require both approaches. An NOV3 application may interwork 
with a network domain that is not implemented by NVO3 technology. For example, 
DC application in a DC provider site provides a service for an enterprise 
company who are on an physical infrastructure. The first approach is required 
to be used here. 

The second approach is useful for many applications. IMO: NVO3 WG has not 
worked on it from the architecture perspective. The third domain (your word, or 
overlay network) architecture should have edge node and transit node components 
and aim to use tunnels as transport link between any pair of nodes in an 
overlay network (third domain). Current NVO3 arch is for an NVO3 network with 
edge nodes only, i.e. single hop from NVO3 network perspective. 

Thanks,
Lucy
        
-----Original Message-----
From: Joe Touch [mailto:[email protected]] 
Sent: Monday, October 10, 2016 9:55 AM
To: Greg Mirsky
Cc: Kiran.Makhijani; Anoop Ghanwani; Sam Aldrin; Bocci, Matthew (Nokia - GB); 
NVO3; Lucy yong; Alia Atlas
Subject: Re: [nvo3] Discussion on encapsulation formats and next steps



On 10/8/2016 10:43 AM, Greg Mirsky wrote:
> Hi Joe,
> I think that he task of OAM interworking between two NVO3 domains that 
> use different dataplane encapsulations would be much simpler if their 
> respective OAM are compliant to the same comprehensive set of 
> requirements.

As I noted, there are two approaches:
1) dovetail the two domains and translate between them, which supports the 
subset of the capabilities of the two domains and requires translation.
2) layer a third domain over the two domains and treat the two as independent, 
relying on the common third domain to establish capabilities

This is exactly the difference between pre-Internet internetworking and 
Internet internetworking. It's been 47 years since we tried the Internet 
experiment; it surprises me we haven't concluded whether it worked or not yet, 
but if this WG wants to try the pre-Internet approach one more time to see if 
it works, so be it.

Joe
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to