> -----邮件原件----- > 发件人: [email protected] [mailto:[email protected]] 代表 > Thomas Narten > 发送时间: 2012年6月30日 5:48 > 收件人: Dhruv Dhody > 抄送: [email protected] > 主题: [nvo3] Oracle Terminology [was: Re: call for adoption: > draft-narten-nvo3-overlay-problem-statement-02] > > Dhruv Dhody <[email protected]> writes: > > > (2) Sec 3.3 > > "The first component consists of a back-end "oracle" that > > is responsible for distributing and maintaining the mapping > > information for the entire overlay system." > > > Should we use the word "oracle" in the ID? IMHO using a more generic term > > would be much better. > > I don't particularly like the term "oracle" myself, but that was the > best I was able to come up with. Other suggestions welcome. > > Note: the whole point of using a term like "oracle" is that it says > nothing about how the oracle itself is implemented. As others have > mentioned, other terms (like "directory based") have "bagggage" > associated with them that could be read to imply particular solution > approach. > > The oracle itself could be implemented as a directory or database of > some sort (single instance, distributed, or whatever). > > Or it could be implemented via an existing (or modified) routing > protocol (BGP, IS-IS, etc.). > > Or it could be implemented as part of the orchestration systems used > to manage/migrate VMs. > > Because there are a range of plausible approaches for implementing the > oracle, it seems desirable to use a standardized protocol for the > NVE-oracle interaction. The NVE can then use a single protocol to > query the oracle for the mappings it doesn't have, or to push mappings > to the oracle for any VM's the NVE supports without caring how the > oracle itself is implemented/architected.
Fully agree that it would be much better to use a standardized protocol for the NVE-oracle interaction if appropriate. As an example of using BGP, a standardized routing protocol for the NVE-oracle interaction, section 3.6.3. "RIB Reduction on PE Routers" of Virtual Subnet (http://tools.ietf.org/html/draft-xu-virtual-subnet-07) briefly explains how to use BGP tricks to push and query mappings between the NVE (i.e., L3VPN PE) and the oracle (i.e., Route Reflector) . Note that one of the advantage of using BGP as NVE-oracle interaction protocol is that the mappings cached on the NVEs (e.g., L3VPN PEs) could be updated in a timely fashion, which has been listed as one of the basic NVo3 control plane characteristics in the NVo3 control-plane requirement doc. Best regards, Xiaohu > If we have a separate protocol, we aren't tying the NVE to a > particular implementation/approach for the oracle. > > 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
