I am considering LISP implementation within two Data Centres and a mid-size Campus (triangle topology) but I need some answers & shall appreciated your inputs.
[DC1]====[DC2]
\ /
\ /
[CAMPUS]
Classic MACSec(CTS) + OTV on DCI links (all DCI links are IP over DWDM). vPCs on Core and Access layer. Also, it will be a multi-tenant environment so we consider MPLS + MPBGP (classic L3 VPNs) across the whole network.Multicast should be also considered, but only P2MP (e.g. no MP2MP) so we consider RSVP-TE P2MP TEs + BGP Auto discovery. I am giving these details as it may affect the LISP design as well.
The reason why we consider LISP is because of Wireless LAN controller operation and need of stretching L2 VLANs of WIFI networks across both DCs. So I decided to think about step or two further and use LISP for few other things (such as Load Balancing and VM Mobility).
Now to the point. My questions are as follow:
1. Is there any document that suggests how it’ll be best to implement LISP in conjunction with other routing protocols? At the moment I am tempted to go with the following design: OSPF Multi Area, where DC1, DC2 and Campus network are in different (probably stub) areas. Then all other links being LISP enabled. Does that sound right?
2. Are there any gotchas with LISP and vPC interoperability? Is it supported to have FHRP on LISP enabled Interface on N7K?
3.As LISP is Map-and-Encap, meaning that all the traffic between “PE”s can be encapsulated, is it a good idea to get rid of MPLS + MPBGP and stick with LISP only? Since LISP is natively capable of handling multi-vrf topologies, is that a good idea?
4.Will I need OTV at all, if I go with LISP only? I cannot think of any reason why would I need it ..
5. I haven't read much about LISP and Multicast, VRF enabled environments. Are there any gotchas there?
6. Is it better idea to use LISP for extended networks only and stick with the classic IGP+MPLS+MP-BGP implementation for the rest of the network?
Some constraints that I can think of of the top of my head are:
1. If I make the whole network LISP enabled, I will fix my inefficient routing issues, but I will probably loose visibility for traffic flows (we’re having some intelligent TAPs in the network).
2. Firewalls (We’re using PaloAlto) doesn’t know anything about LISP, yet..
3. Supportability will be a bit challenging due to tech team unawareness of this “new” architecture (see, I didn't say protocol ;)).
Apologies for the long topic & many thanks in advance for your time! Any thoughts will be more than appreciated!
D.
