Thomas, maybe to avoid the risk of misunderstanding, as the term "collocation" could mean: - in the same physical building or in the physical room (that is the meaning of co-location in the telco world, when a CLEC co-locates at a LEC building). - on the same physical server, but at the same level as the hypervisor, which means in case there are 2 hypervisors configured on the server, one NVE may serve both. - on top of the hypervisor/ as a part of the hypervisor/ integrated with the server hypervisor/ i.e. the NVE is kind of a plug-in to the hypervisor or to an V-Switch that is controlled by the hypervisor. This means, that if there are 2 hypervisors installed on the Server, than each must have its own NVE.
I think that either one or both last meanings are meant, but it may be could to make it more explicit what is meant. Lothar -----Ursprüngliche Nachricht----- Von: [email protected] [mailto:[email protected]] Im Auftrag von Thomas Narten Gesendet: Donnerstag, 17. Oktober 2013 22:55 An: Lizhong Jin Cc: [email protected] Betreff: Re: [nvo3] NVO3 Architecture document Lizhong, Lizhong Jin <[email protected]> writes: > Hi Thomas, > In section 4, you describe two NVE models. The model name "NVE Co-located > With Server Hypervisor" is not so accurate, and I suggest to change to > "NVE Co-located With Server" or "NVE integrated in Server". In the > implementations, part of NVE function will be performed by the adaptor, > not hypervisor. The function performed by adaptor will be more than > checksum offload and LRO/TSO, it is even possible in the future that the > full NVE is embedded in adaptor. I think it is important in the wording at issue above to include the word "hypervisor" because Section 4.1 specifically addresses the case of server virtualization. This is in contrast to Section 5.1, which talks about the case of Network Service Appliances, where there is no hypervisor, but the NVE is also implented on the server. With that in mind, I'm a bit bit puzzled by the request. I don't think it really changes Section 4.1 in any way. That makes me wonder if I'm not understanding the real concern behind your request. The second paragraph of Section 4.1 specifically talks about adaptor offloads, making it clear that some functionality may be offloaded onto the adaptor for performance reasons. If your concern is really about properly addressing adaptor offloads, is there specific wording in the second paragraph that you think needs tweaking? Thanks! Thomas _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3 _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
