IMO: "over the network layer" is another important aspect for NVO3. We want to decouple application networking from the infrastructure network.
Thus, Virtualization and Overlay are the key.
Lucy
From: Eric Gray [mailto:***@ericsson.com]
Sent: Tuesday, June 04, 2013 3:53 PM
To: Linda Dunbar; ***@ietf.org
Cc: Lucy yong
Subject: RE: [nvo3] Is there any difference between "Network Virtualization Overlay" and "Network Overlay"?
Linda,
I could be wrong, but I think we're confusing the "scope" of the NVO3 working group with
the work that might actually get done in NVO3.
First, we need to remember that NVO3 is not currently chartered to define any solution.
It is chartered to "consider approaches to multi-tenancy that reside at the network layer"
- which certainly includes L3VPN and likely includes aspects of L2VPN that have to do with
network virtualization "at the network layer" (i.e. - over layer 3, which is how we do it).
This latter point was discussed months ago. Because we defined L2VPN, Pseudowires,
etc. here (in the IETF), if we now discover that there are compatibility (or other) issues
with using these approaches as a network virtualization technique, we have to fix those
issues here.
It is certainly within the scope of the charter for NVO3 to evaluate this.
That does not mean that the NVO3 working group would take a direct active role in fixing
issues we find with either L2VPN or L3 VPN. As I said, fixing issues isn't in NVO3's charter
in any case and fixing anything that belongs to another IETF working group is an activity
that is either unlikely in NVO3, or would be explicitly delegated to NVO3 at some future
time - as part of a re-chartering evolution.
It is very odd to try to talk now about the "scope" of a potentially re-chartered NVO3 WG
at some future date.
As for MPLS, I believe that the IETF once had a broad discussion about whether or not it
is a Network Layer thing. I believe the conclusion was that MPLS is closely tied to the
network layer (and IP in particular).
In addition to L2VPN and L3VPN as applications of MPLS, there is also Traffic Engineering.
I don't believe we have yet eliminated the potential need for traffic engineered paths in
virtual networks used for data-centers. And I doubt anyone is willing to consider using
IP explicit routing to accomplish this, in the even that it is needed.
If we limit the scope of what NVO3 needs to consider to strictly IP encapsulation we may
be dodging many the problems we've talked about in NVO3. In many cases, that might
be okay, but we don't know that at the moment...
--
Eric
From: nvo3-***@ietf.org [mailto:nvo3-***@ietf.org] On Behalf Of Linda Dunbar
Sent: Tuesday, June 04, 2013 4:10 PM
To: Lucy yong; Eric Gray; Larry Kreeger (kreeger); Thomas Narten; ***@hobbesdutt.com; Black, David; Murari Sridharan; ***@ietf.org
Subject: Re: [nvo3] Is there any difference between "Network Virtualization Overlay" and "Network Overlay"?
Importance: High
Snip...
. When MPLS is not available, using mpls in GRE encapsulation, L2VPN/L3VPN can be over an IP network, which fits in the NVO3 scope.
[Linda] That is exactly what I want to say: i.e. L2VPN/L3VPN achieved via IP encapsulation (GRE, VxLAN, etc) is in the scope of NVO3. The L2VPN/L3VPN achieved via MPLS encapsulation fits in the scope of L2VPN WG and L3VPN WG respectively.
Linda