Hi Marc, Nabil, Thank you for the reply. I do not mean specific solution, but hope to see more specific requirement about VNID, which would be helpful for future solution. Current requirement is very general to include all possible solutions. I could provide two points: 1. do we require any different kinds of control plane to interoperate for NVO3? 2. do we require dynamic control plane and static configuration to be interoperate for NVO3? And yes, a bit related with control plane requirement. If we could answer the above points, that would be helpful to refine VNID requirement.
Regards Lizhong 2012/5/24 LASSERRE, MARC (MARC) <[email protected]> > Hi Lizhong,**** > > ** ** > > This is a requirements draft and not a solutions draft where a detailed > VNID format needs to be specified.**** > > The intent of the original sentence implies that either a global > identifier or a local identifier (such as an MPLS label) can be used.**** > > ** ** > > Are you asking for a specific change in wording – i.e. such as mentioning > that a local identifier is typically exchanged via a control plane protocol? > **** > > It is pretty implicit and since it is a data plane requirements draft, we > did not want to get into control plane related aspects. **** > > If clarification is needed, we will be happy to be a bit more specific.*** > * > > ** ** > > Marc**** > > ** ** > ------------------------------ > > *From:* Lizhong Jin [mailto:[email protected]] > *Sent:* Wednesday, May 23, 2012 4:29 PM > *To:* LASSERRE, MARC (MARC); [email protected] > *Cc:* [email protected] > *Subject:* Re: New dataplane requirements draft**** > > ** ** > > Hi Marc, Nabil,**** > > In draft section 3.3.1.1, it describes the functions of VNID:**** > > "A VNID MUST be included in the overlay encapsulation header on the > ingress NVE when encapsulating a tenant Ethernet frame or IP packet on the > overlay tunnel. The egress NVE uses the VNID to identify the VN context in > which the encapsulated frame should be processed. The VNID can be either a > globally unique identifier (on a per-administrative domain) or a locally > significant identifier. It MUST be easily parsed and processed by the data > path and MAY be distributable by a control-plane or configured via a > management plane."**** > > > It does not implicit any format of the VNID. There are potentially two > formats, one is like MPLS lable allocation which is implicit for dataplane > and mapping with control plane information, the other is like VLAN format, > which is explicit for dataplane and could work without control plane. In > http://www.ietf.org/id/draft-kj-nvo3-pion-architecture-00.txt, section > 5.3, it describes this. We should gather some requirement about this, and > hope to see more comments.**** > > **** > > Lizhong**** >
_______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
