[c-nsp] ISP / MPLS "POP" design

[Oliver Boehmer (oboehmer)]

>
>> no doubt here.. hence I really wonder who would ever put
>> forward such a requirement (and Adam added a smiley, so
>> not sure ;-).. We have unified-mpls to scale to very
>> large MPLS domains, but the IGP certainly doesn't need
>> to scale anywhere close to this..
>
>RFC 3107, I'm assuming.

yep, along with multi-area/level/domain IGP..

>
>Even BGP-LS wouldn't necessarily scale, as it's exported
>data from the IGP (which won't scale to millions of routes
>anyway).

Hmm, I consider BGP-LS only to distribute existing topology information to
"clients" (for example some controllers or PCE or whatever is interested
to learn about the network topology), not to update the RIB/FIB.

>
>Even the data centres are turning to BGP to scale east-west
>routing away from the IGP's (draft-lapukhov-bgp-routing-
>large-dc-06).

of course, what else? ;-)

	oli
>