[c-nsp] VPLS, spanning tree and redundancy

[Pshem Kowalczyk]

On 25/03/2008, Enno Rey <erey at ernw.de> wrote:
> Hi,
>
>  just a question in advance: do you know for sure that STP BPDUs are forwarded "across the cloud"?
>  [this may or may not be the case, depending on the gear used and the configuration/carrier (you?) providing the VPLS links]

We're providing the VPLS links and the BPDUs are definitely forwarded
(we have some customers with 'hub and spoke' design that wanted all
their traffic to go through a central point and STP solved that
problem for them.
> if I understood you correctly there will be two different "VPLS clouds" between both pairs of CSW/PE links?
>  why run STP at all, then?

Yes, the idea is to run two VPLS clouds, we wouldn't run the STP, but
the customer will.
> again: given two different clouds this shouldn't happen.

Unless you use spanning tree and cost of going directly between
switches is higher then going all the way around. We don't have
control over the CSW - so it's up to customer to figure that out :-)

>
>  furthermore I seem to recall a Cisco white paper discussing ways (though impractical ones, from my perspective) to avoid loops in VPLS scenarios:
>  http://www.cisco.com/en/US/tech/tk436/tk891/technologies_white_paper09186a00801f6084.shtml
>

Thx for the link, will have a look.

kind regards
Pshem