Linda,

Linda Dunbar <[email protected]> writes:

> Larry,
> 
> Could you please explain the following question?
> 
> LK> I don't agree that the NVA should be connected to an NVE.  I see
> no benefit to this.  The NVA should connect directly to the
> underlay, just as the NVEs do.   
>  
> [Linda] NVA has to be placed somewhere in a data center. Say a NVA
> is placed on a server rack. Most server racks are pre-wired with
> each slot to the ToR switch. In an environment where the ToR is the
> NVE, then the NVA is attached to the NVE.  How would you call this
> attachment? 

Maybe I misunderstand, but I read the above as saying that the NVA
could be on its own VN and thus reachable only via an NVE. If that is
what you mean, I have to wonder why we want to do that. Some
considerations:

1) It would seem to raise recursion problems... An NVE that needs to
get address mappings attempts to contact the NVA, but it needs to
first get address mappings for the VN to which the NVA is connected, oops...
All this is avoided by having the NVA connect to the underlay and be
accessible via underlay addresses.

2) Why would you want to do this? What benefit is there? I don't see
any advantages right off.

Thomas

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

Reply via email to