> > A subnet (L2 VPN) could communicate with end-systems on other subnets > > via a default gateway IP interface that is in a VRF which is a member > > of an L3 VPN. > > And, right on cue, Aldrin tells us why IRB is needed :-) > > Totally agreed: necessary, whether one likes it or not. To channel > Maria, though, one could "simulate" this by IP routing both within and > between subnets. A uniform forwarding approach would make things a lot > simpler. So the question comes back: why consider a "subnet" a Layer 2 > entity?
Precisely. Hence, the only question that remains is how to address non-IP traffic, for those DCs that care about such traffic (where non-IP traffic is the actual application traffic, not DHCP or ARP generated by the guest OS). It would seem that the best way to answer that question would be to first to measure such traffic. Maria _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
