And since all of this discussion is in solution space, the answer is not to use such middleboxes on the WAN side of the DCBR when asymmetric routing is possible on the WAN side - they work fine on the data center side. The meta-point is that this approach is viable and excluded by the language you're proposing to add to this draft - that's not appropriate for a general issues draft, but could be appropriate for a solution architecture draft for the specific sort of solution that the vm-issues draft appears to envision.
Thanks, --David From: Linda Dunbar [mailto:[email protected]] Sent: Tuesday, September 23, 2014 6:37 PM To: Black, David; Larry Kreeger (kreeger); [email protected] Subject: RE: [nvo3] WG Last Call on draft-ietf-nvo3-vm-mobility-issues David, Is your "hot potato" routing referring to outbound traffic going through a different DC Gateway than the inbound traffic? Then there will be issues with some middleware boxes that monitor both directions of traffic. Linda From: Black, David [mailto:[email protected]] Sent: Monday, September 22, 2014 7:06 PM To: Linda Dunbar; Larry Kreeger (kreeger); [email protected]<mailto:[email protected]> Subject: RE: [nvo3] WG Last Call on draft-ietf-nvo3-vm-mobility-issues Linda, > In "vm-mobility" draft, the "Optimal routing" is to avoid "triangular > routing" (or what you stated "hot potato" route). I agree on avoiding "triangular routing", but "hot potato" routing is different from "triangular routing" and may be "optimal" depending on how "optimal" is defined. In general, "hot potato" routing refers to connectivity between two networks in which the upstream network hands off traffic to the downstream network as quickly as it can in both directions, resulting in asymmetric routes. This can and does work in practice when used and it would be incorrect to characterize it as something that should be avoided in all cases.
_______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
