> With this so called "control protocol", the packets will be dropped at the 
> Ingress NVE when egress NVE is not reachable.  

Only if *all* the NVEs for the EID? (what do you guys call the things behind 
the NVE) are unreachable.

> Without this so called "control protocol, the packets will be dropped 
> somewhere in the network when the egress NVE is not reachable.  

Right.

> In data center environment where most communications are among applications, 
> most likely a source will not send more packets without acknowledgment from 
> the destination. Then the impact of where the packet is dropped is not that 
> big. 

You have to make this work for UDP applications.

> But in an environment where NVEs have aggregated traffic from many 
> sources/flows, then it is important to have this "control protocol". In this 
> environment, it might even warrant a control plane for NVE to notify the 
> source when the egress is not reachable, so the source can choose a different 
> ingress NVE. 

No comment.  ;-) But that works naturally in MPTCP by the way.

> Bottom-line: I truly think that NVO3 needs a better boundary on the problem 
> domain. If NVO3 is targeted for solving problems for entire worldwide 
> internet, you need many control plane protocols. If NVO3 is targeted for 
> solving problems in data centers, then many control plane protocols becomes 
> unnecessary. 

Dino

_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to