[c-nsp] BGP-ORR vRR/IOS-XRv

[Curtis Piehler]

I realize my previous post about MCID across a RR backbone generated
various difference in opinions so hopefully this topic is a bit more cut
and dry.

BGP-ORR looks to be an attractive technology to utilize instead of full
meshing many standard RRs.  I was hoping people could share some insight on
this topic since there is not so much real world scenarios out there
documented.  This is geared towards an IOS-XRv ASR9K deployment.


   - When running BGP-ORR a vRR computes the shortest path for a given
   BGP-ORR enabled client from the perceptive of the client utilizing the root
   node of a BGP-ORR
   - For this reason is it suggested that BGP-ORR groups are configured
   with respect to common views of the network (IE: grouped by POP/Region)
   instead of having one or two ORR Groups?
   - What is the suggested max number of BGP-ORR groups per vRR?
   - What is the suggested max number of clients per BGP-ORR group?
   - If all clients in the network peer with say 2 vRR as the center will
   those clients only see maximum two BGP paths (This is assuming add-path is
   not utilized)
   - Do all clients need to peer with all vRRs as long as the vRRs are
   fully meshed?
   - Can anyone share general experience they had with BGP-ORR deployed in
   a production backbone?
   - Are there any substantial gotchas to be concerned with?


As per Cisco documentation when running OSPF a basic MPLS-TE configuration
is required on all root nodes, correct?

Specific to IOS-XRv does anyone know how the new "SUB" licence models
work?  I have not been able to find Cisco documentation that explains
this.  The perpetual/term licence model goes End of Sale Q3 of this year.

https://www.cisco.com/c/en/us/td/docs/routers/virtual-routers/xrv9k-62x/general/release/notes/b-release-notes-xrv9k-623.html

Thanks
Curtis