EIGRP to IS-IS... Really?

BeTheNetwork

I am a consultant and I have a client that is changing out 4 6500's for 2 Nexus 7k's. The problem is their IGP is EIGRP and only recently did Cisco up the number of EIGRP adjacencies on the NX-OS platform to more than 50. This customer has 400 remote-site locations that will attach to these core devices. The remote-sites are interconnected via a SP MOE service.

My first thought is to use BGP, but the customer came back and said that they want to use IS-IS, because of the BGP learning curve. Ok. My first thought is a single-area flat level2 domain and verify that CLNS can pass through the SP transport.

Any thoughts on other considerations?

Thanks,
Brent

networker050184

So they aren't worried about eh learning curve with IS-IS?

Anywho, yeah I'd go with a flat L2 and and should be good to go there. My bigger concern would be 400 IGP adjacencies!

rakem

400 remote sites?
BGP is designed for that type of stuff. Sounds like you need to school them a bit.


How do they all connect in? I'm guessing via some type of MPLS service?

cisco_trooper

**** wrote: »

My first thought is to use BGP, but the customer came back and said that they want to use IS-IS, because of the BGP learning curve. Ok.

Sounds like they must already know all about ISIS then...

BeTheNetwork

I have to do some discovery. MOE implementation is hit and miss. I have seen q-in-q in some cases. A dedicated L3VPN to the customer would be more difficult, IMO. I've never used integrated is-is as a PE-CE protocol. I don't know if its even possible - you'd think so given the nature of MPLS transport. BGP is the way to go in any case. The problem there becomes i vs.eBGP operation, but I can work through that.

I did some looking on ipspace.net yesterday and found the following...



If I read that right, I don't need to worry about CLNS in a Q-in-Q because IS-IS has its own Ethertype (0xFE), just as long as the SP doesn't filter it, it may work. I will have to noodle through and see what works.

Thanks.

powmia

You can wrap IS-IS in a single vlan tag, and you can wrap it in two. Shouldn't be a problem. Still a poor choice of protocols. I would take into account the flooding considerations... possibly silencing the headend (mesh-group blocked) and static defaults on the remotes if you can get away with it.