Route (IP traffic) within subnet is happening, regardless physical or
virtual, vpn or non-vpn, call it enhanced/advanced forwarding.
Luyuan

-----Original Message-----
From: Lucy yong <[email protected]>
Date: Tuesday, December 18, 2012 10:10 AM
To: Luyuan Fang <[email protected]>, Kireeti Kompella
<[email protected]>, "NAPIERALA, MARIA H" <[email protected]>
Cc: Aldrin Isaac <[email protected]>, "[email protected]" <[email protected]>
Subject: RE: [nvo3] FW: New Version Notification for
draft-yong-nvo3-frwk-dpreq-addition-00.txt

>My comments are inline.
>
>> -----Original Message-----
>> From: Luyuan Fang (lufang) [mailto:[email protected]]
>> Sent: Tuesday, December 18, 2012 12:04 AM
>> To: Kireeti Kompella; NAPIERALA, MARIA H
>> Cc: Aldrin Isaac; Lucy yong; [email protected]
>> Subject: Re: [nvo3] FW: New Version Notification for draft-yong-nvo3-
>> frwk-dpreq-addition-00.txt
>> 
>> 
>> 
>> On 12/17/12 1:57 PM, "Kireeti Kompella" <[email protected]>
>> wrote:
>> 
>> >For IP traffic, whether intra and inter-subnet, IP VPNs suffice.
>> 
>> Yep.
>[Lucy] not sure. We are working on the network virtualization. The one of
>goals is to virtualizes the enterprise physical networks that provides
>the networking for compute resource on a DC common infrastructure. If you
>look at today's enterprise data center networks, do they all use routers
>without any switch? Not at all. If routers can do all the functions, why
>it is not the case today in any physical network? You may argue that they
>are legacy. But I do not see a new built DC without a switch too. Why is
>that? How can we assume that IP VPN is sufficient for a tenant virtual
>network.
> 
>> >
>> >The solution is simple: route if IP, bridge if not.  Yes, one could do
>> >IRB, but why?  IRB brings in complications, especially for multicast.
>> 
>> Exactly, thank you for speaking out.
>[Lucy] Again, IRB are widely used in enterprise data center. It is
>useful. How do we know that IP VPN is not complicated for a tenant
>virtual network? 
>> 
>> 
>> >I'm sure someone suggested this already, so put me down as supporting
>> >this view.
>> 
>> I believe Pedro gave detailed explanations many times a few months ago.
>> Of
>> course, Maria and I had been expressing such opinions all along on-line
>> and off-line, and brushed off good amount of bullets. :-).
>[Lucy] IMO: Pedro's solution applies to certain application, but not
>other application. For example, his draft does not mention how to provide
>a broadcast for a subnet. Unless a tenant virtual network does not need
>this at all, we can not remove this requirement.
>> 
>> >
>> >If there is a case for _only_ intra-subnet traffic, one may create an
>> >E-VPN to handle both IP and non-IP; but I suspect this is a rare case.
>> 
>> A solution just for some very narrow cases is not that too useful.
>[Lucy] Not at all. Another objective of NV03 is the ability to move TS
>among different servers regardless if they are on the same or different
>subnet. Do you see this is "very narrow cases"?
>
>
>Sorry, we are back to use VPN terms again. I did not start this.
>
>Regards,
>Lucy
>> 
>> >From that point of view, I would like to see E-VPNs in the data center
>> >*always* coupled with IP VPNs, and only dealing with non-IP traffic.
>> 
>> Do you mean the following? IP VPN alone is sufficient if all traffic is
>> L3
>> or handling the majority traffic which is IP; but E-VPN is not in the
>> same
>> case, E-VPN needs to leverage IP VPN for inter-subnet routing (unless
>> you
>> never get out of a subnet), therefore it needs to "*always* coupled
>> with
>> IP VPNs"?
>> If this is correct understanding, and we are talking about VPN cases
>> only
>> (there are also other non-VPN options too), I agree with you.
>> 
>> The "*always* coupled" wording makes me a little nervous, better to ask
>> you to elaborate.
>> 
>> What I don't want to see is that the implementation ties IP VPN and E-
>> VPN
>> together. If the traffic is IP (might well be 90+% cases), there should
>> not be any E-VPN involvement. If both L2 and L3 are needed, there
>> should
>> be clear demarcation, not mixed together, not sharing the same VRF, not
>> make one VPN with some l2 sites and some l3 sites. Let's keep it clean
>> and
>> simple.
>> 
>> >
>> >This may appear drastic, but I think operationally, this is will
>> simplify
>> >things.
>> 
>> Yes, operation simplicity is the key to success. IP VPN is a living
>> proof.
>> Others are not at par.
>> 
>> I think Thomas Narten and Maria both expressed the point to avoid
>> complication, agree with them too.
>> 
>> 
>> I have a larger question, not to Kireeti or any particular person.
>> How come we only talk about MPLS VPNs here now?
>> What happened to other technologies? Were not VXLAN and NVGRE the
>> original
>> motivation for nvo3?
>> I am a MPLS VPN person, but I also don't believe in one size fits all.
>> IP
>> VPN or E-VPN are right for some operators, cannot say for everyone.
>> Where
>> are the folks talking non MPLS topics?
>> 
>> Luyuan
>> 
>> 
>> 
>

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

Reply via email to