[c-nsp] Spanning-tree boundaries and configuration

[Michele Bergonzoni]

Peter Rathlev ha scritto:
>> Should I participate the customer STP in my topology for preventing
>> loops and broadcast storms or ignoring his BPDU's totally?
> 
...

> Otherwise there's AFAICT no way around joining the STP domain of your
> customers. In that case you should of course try to make sure that L2
> problems coming in from one customer doesn't affect other customers.
> 

I think we all agree that joining the STP domain of the customer can be 
a major nightmare, and using pseudowires or enabling bpdufilter is the 
way to go for non redundant connections. I've never seen an MST multi 
region deployment in a carrier/customer scenario (I'm not even sure that 
MST was designed for that).

For redundant connections, VPLS is the technology specifically designed 
for this environment. If it's not available, what can be done is to 
deploy two separate redundant circuits (or two separate redundant metro 
VLANs, if there are more than two endpoints), and let the customer use 
its STP of choice to avoid loops. You transport customer STP as any 
other customer traffic, with "l2protocol-tunnel". This is not perfect, 
but mostly works.

Regards,
                            Bergonz