[c-nsp] IS-IS as PE-CE protocol

[Radu-Adrian FEURDEAN]

On Thu, Mar 21, 2019, at 16:45, Victor Sudakov wrote:
> Well, the Internet is full of examples and recommendations of an IGP
> (most often OSPF) being used between PE and CE, so it must be common
> practice. In fact, OSPF even has special enhancements for this very
> purpose.

It is not uncommon, but it is well-hated.
To start, on the PE you have to do it as a separate instance in the client's VRF (not in the main table, not in a "common" internet VRF), or you will shoot yourself in the head, not in the foot.
Then things may get a little more relaxed if there is a "provider-managed CPE", between the client-owned equipment and the PE.

The main problem is that  there is (??? was ???) generally no route-filtering on IGPs, compared to BGP, which severely limits what a provider can afford to accept.

> Again, it seems a common practice to run an IGP with a customer to
> import the customer's routes into the provider's MPLS network.

Provider-wise, BGP is common. On our side, we use BGP rarely but we prefer it if a routing protocol needs to be run with the customer. We usually use "subscriber routes" (static routes triggered when the subscriber session goes up, removed when it goes down). Then BGP (bigger customers that can afford it), otherwise RIP (yes, RIP !!! - very good "route presence protocol") for big enough customers that cannot humanly afford operating BGP. OSPF in about 3 legacy cases, and we are actively trying to get rid of it. Everything imported into MP-BGP.

At the end it all comes down to what kind of clients you have. I'd say that a customer running IS-IS internally should be litterate enough to run BGP, so no reason to run IS-IS with the customer. this is not necessary the case with OSPF.

-- 
R.-A. Feurdean