[a-nsp] Arista and EVPN-VPWS, RFC 8214

[Joris Claassen]

Hi Gert,

On Mon, Nov 11, 2019 at 8:53 PM Gert Doering <gert at greenie.muc.de> wrote:

> Hi,
>
> On Thu, Nov 07, 2019 at 06:08:46PM +0100, Joris Claassen wrote:
> > As no-one else chimed in before, let me do so - even though I am a bit
> late.
>
> Thanks :-)
>
> [..]
> > As per version 4.22.0F Arista also supports Layer 2 Control Plane
> > Forwarding on a selection of platforms (7020, 7280 and 7500 series).
> [..]
> > https://eos.arista.com/eos-4-22-0f/l2-protocol-forwarding/
>
> Are the other platforms not supported "because implementations has
> not been done yet" or because there is a fundamental incompatibility
> with the switch chips used?  (We use mostly T2+/T3 based boxes for
> the customer connections, that is, 7050SX2 and 7050SX3).
>

>From what I understand this is functionality that is not scheduled to
arrive on T2+/T3 any time soon. As you might know these are part of an
entire different chipset family then the 7020/7280/7500/7800 series on
which the development effort for this feature has focussed.


>
> > MAC learning can be turned off on a per-port basis, as per the manual:
> >
> > https://www.arista.com/en/um-eos/eos-section-19-3-mac-address-table
> >
> > This combination of features will:
> >
> >    - Use EVPN to fill the flood-lists on the respective VTEPs that host
> the
> >    endpoints of your VPWS service
> >    - Forward tagged/untagged/all type PAUSE/LACP/LLDP/MACSEC/STP frames
> >    through the VXLAN tunnel
> >    - Not learn endpoint MAC addresses on the ethernet ports that you
> would
> >    map to the VPWS service
> >    - All the traffic will be transparently sent to the other side as the
> >    MAC address will not be learned
>
> Ignoring L2 protocol forwarding for the moment, I'm a bit confused on
> how to implement the VPWS part.  Is it just an EVPN L2 VPN with "just two
> member devices", and the rest happens quasi automatically, because "without
> learning, all you can do is flooding, and if you only have two endpoints,
> this is a point-to-point pseudowire"?
>

That is correct. All traffic will be classified as BUM and thus be flooded
to all other (only one in this case) VTEPs that are known for that VNI.


>
> gert
> --
> "If was one thing all people took for granted, was conviction that if you
>  feed honest figures into a computer, honest figures come out. Never
> doubted
>  it myself till I met a computer with a sense of humor."
>                              Robert A. Heinlein, The Moon is a Harsh
> Mistress
>
> Gert Doering - Munich, Germany
> gert at greenie.muc.de
>

Best regards,

-- 

<https://www.arista.com/>
Software Driven Cloud Networking

Joris Claassen
Systems Engineer EMEA
joris at arista.com
+31 (0) 648538869
*www.arista.com <http://www.arista.com/>*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/arista-nsp/attachments/20191112/0185d161/attachment.htm>