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

[Mark Tinka]

On Thursday, November 07, 2013 11:38:22 AM Adam Vitkovsky 
wrote:

> Yup RFC 3107 is what I proposed but I guess some of the
> folks would try just about anything to maintain the
> status quo instead of implementing separate IGP domains
> and hierarchical MPLS across the globe and I certainly
> understand why, I mean this could be a carrier changing
> project if not executed seamlessly.

Indeed.

RFC 3107 is not without its implementation issues, 
particularly in a multi-vendor network.

I tested it once in a Juniper-only implementation, and that 
was hairy enough. But it does work.

> Inter-AS MPLS traffic engineering works just fine so the
> need for MPLS TE is not a reason to avoid RFC 3107 in
> Inter-AS deployments.

I was referring to Inter-Area TE, rather than Inter-AS TE, 
as being an issue where RFC 3107 introduces breakage due to 
scaling the IGP through the use of Areas and Levels.

But, PCE orchestration or SR are possible solutions here.

> I can definitely see BGP-LS maintaining LSDB for several
> millions of prefixes replacing ISIS and OSPF altogether
> in ISP cores and large DCs.

I think you mean draft-lapukhov-bgp-routing-large-dc-06, 
since BGP-LS, AFAIK, takes its data from the IGP. So it will 
only scale as much as the IGP will scale.

BGP-LS, if you think about it, is a perverse method of IGP-
BGP redistribution. Interesting how technology always 
disguises itself and comes back :-).

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20131107/8a374b5b/attachment.sig>