[c-nsp] question about service provider network design

[Dan Armstrong]

So say I have an SVI on a PE switch which in turn has 2 layer2 links 
back to 2 core boxes, the core boxes protected again by a 3rd layer2 link.

MST will protect me and make sure I always have link to the PE routers & 
core routers.  What's wrong with using that SVI address in your PE 
router as a reference, no need for an IGP down there?





Nathan wrote:
> On Tue, Oct 21, 2008 at 4:16 PM, Dan Armstrong <dan at beanfield.com> wrote:
>   
>> We have a fairly similar design for our Metro Ethernet network.
>>
>> Our primary method of protection is STP(MST).  I've been thinking about
>> this, and I can't come up with a reason why we even really need an IGP down
>> to the edge PE devices?  Since it's all layer2 - the core switch/routers see
>> all of the PE<>core links as Connected routes anyway - what's the point of
>> bother pushing your IGP down there? It's just more needless routes.
>> That leaves you with a very small IGP in your core.
>>     
>
> The problem is that you are supposed to have redundant links between
> routers. The way to have permanent links between routeurs in spite of
> changing routes and falling interfaces is to establish communication
> between loopbacks, and that is what LDP and iBGP - MPBGP do. Therefore
> you need unfettered communication between the loopbacks of your
> routers, PE routers included, therefore you need your loopbacks in
> your IGP, therefore you need IGP on your PE routers.
>
> I suppose you could somehow make the network function without it, but
> you'd lose redundancy at the very least.
>
>