[c-nsp] VPLS, spanning tree and redundancy

[Pshem Kowalczyk]

Hi,

It looks like one of our customers would like to have redundant L2
access. We have two PEs in both locations so in theory that should
work. However, I wonder what your  ideas are about preventing L2 loops
in such network. The customer envisaged something like this:
(variable-width font):

CSW1----CSW2
   |            |
PE1       PE2

 MPLS cloud

PE3        PE4
   |             |
CSW3---CSW4

where CSW is the customer's switch. Currently they only have one half
of that solution (CSW1, PE1, PE3 and CSW3) with a single VPLS between
PE1 and PE3.

The solution we came up with is to run single VPLS1 between PE1 and
PE3 (the way it is now) and then add VPLS2 between PE2 and PE4. This
way customer can run spanning tree (bearing in mind that if they set
the cost incorrectly they will get CSW3 talking to CSW4 through CSW1
and CSW2).

I know that we might have scaling issues in future (twice the number
of VPLSes), but are there any other issues that you're aware of?

kind regards
Pshem