> On Jul 14, 2017, at 7:11 PM, Dale R. Worley <[email protected]> wrote: > > Joe Touch <[email protected]> writes: >> Even a NVI isn't really flow information, so might not have any >> bearing on whether a set of packets (with the same NVI) should maintain >> their relative order. > > Well, the definition of "flow information" is not in the positive > sense -- if the value is the same between two packets, their order must > be maintained -- but rather in the negative sense -- if the value is > *different* between two packets, their order *need not* be maintained. > That causes devices to behave conservatively, in that they are allowed > to reorder packets only if they have specific knowledge that the packets > are "in different flows". From that point of view, if a device knows > that a certain field is an NVI, then it can treat that field as flow > information.
Using Geneve VNI doesn't take us towards that goal, though. On a simple network, there is only one VNI (one virtual network), and if we used solely VNI for underlay ECMP, all tunneled packets would go over the same ECMP path. We want each tunneled TCP flow to take one path, and ideally the next tunneled TCP flow taking another ECMP path. -d _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
