On 4/May/15 11:13, Saku Ytti wrote: > Why not? To me 6PE is easy, if you run MPLS, you want 6PE. After all, it's > nothing but IPv6 in VRF, which I'd want in IPv4 as well, if I can figure out > good brown-field migration to it. > Day1 you get to benefit of all of the MPLS goodness. And in your case, as you > run BGP free IPV4 core, it feels like sentimental choice of not running 6PE. > > Why must there exist coupling between core signalling and edge services? Less > coupling, less complexity.
I find MPLS to be complex, already. So if I could avoid it, I would, but l2vpn, NG-MVPN and TE are good services that are simply easy to deliver over an MPLS network. So I can live with it. I've ran into issues where IPv6 has saved me because we broke IPv4. The ships-in-the-night approach has won out countless times. Losing your MPLS backbone would kill some or all of your IPv4 network. It would be a shame for IPv6 to share the same fate unless it really had to. For me, 6PE falls in the "does not really have to" realm, because native IPv6 is basic... basic enough to work all the time. I'm all for getting the basic right first, and then add complexity afterward - which is why I don't like to teach MPLS to students that do not understand IP. Yes, with a shared LDP, RSVP, SR or IS-IS code base, you do get back to fate sharing to some extent, but vendors have knobs to provide adequate separation where necessary (or if they don't, you ask for them), and by then, we've figured out reasonable operational tricks to work around issues caused by shared fate. But in the early days, especially since MPLS was reasonably new back in the early-to-mid 2000's, I'd prefer to avoid the hard stuff, however easy it looks. Mark. _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
