[c-nsp] Cisco Nexus as MetroE switch?

[Chris Evans]

I would avoid TRILL going forward IMHO.. Most vendors have abandoned future
investment with it.   Plus there are limitations with the Broadcom Trident
asics where you can't decap & route natively (without doing some special
packet looping to workaround this limitation)..

On Mon, Nov 2, 2015 at 3:06 PM, Tom Hill <tom at ninjabadger.net> wrote:

> On 01/11/15 08:04, Marian Ďurkovič wrote:
> > Nowadays some other approaches are also available. For example you can
> use 1RU
> > TRILL switches with 48 * 10G/1G SFP+ and 6 * 40G QSFP+ ports to do
> exactly what
> > you're looking for - without the complexity of MPLS/VPLS or VXLAN+BGP and
> > without ugly workarounds to e.g. achieve proper load-balancing over
> parallel links.
>
> I don't know if you're referring to the part of the SIX slides
> (referenced later in the thread) that said:
>
> "Initial support for lab testing didn't support per-flow load balancing"
>
> But I do know that other vendors whom have had problems with egress link
> load-balancing for VPLS on Trident+, have also had the same issues with
> TRILL (Extreme X670V, in my case) and thus, struggle to suggest TRILL.
>
> But in all honesty, I don't think TRILL suits the use-case of Metro-E
> aggregation - especially where E-LINE services are required (as per the
> OP). To a point, fine-grained Labels would provide an equivalent service
> separation to that of QinQ, but it isn't as segregated as the 'full'
> encapsulation of AToM, VPLS, or PBB/SPB-M, and FGL wasn't possible in
> Trident+ either (T2 only & above, I believe).
>
> The last time I asked a vendor (Avaya) about E-LINE services over an
> SPB-M domain (effectively multipath-capable PBB) some very puzzled
> noises were made. This was ~3-4 years ago now, perhaps they've done
> better, but I learnt not to ask too much of those implementations. :)
>
> Concentrating on the original suggestion, however; it is possible Cisco
> are referring to the use of MP-BGP EVPN for signalling VXLAN instances:
>
>
> http://www.cisco.com/c/en/us/products/collateral/switches/nexus-9000-series-switches/guide-c07-734107.html#_Toc414541674
>
> You can probably get away with that for E-LINE infra, but it's pretty
> complex vs. say, N7K/Cat6800 & AToM. :)
>
> > We have TRILL infrastructure in production for more than one year in the
> Slovak
> > internet exchange (visit www.six.sk for details) as well as in
> Bratislava MAN of
> > SANET.
>
> Even if it was a bit off-topic, I do like that SIX have done something
> differently, and I'm very glad that it works for you - TRILL does
> present a very neat solution for IXPs, especially with regards to
> flexibility in expansion. :)
>
> --
> Tom
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>