Hi David, Flow based refer to forward the traffic based on not only MAC/VLAN(L2) and Dest-IP(L3), but also, e.g, 5 tuple, including TCP/UDP port. This is useful for some network service, e.g, firewall.
Regards Lizhong On Jul 28, 2013 4:39 PM, "Black, David" <[email protected]> wrote: > The question that I have is: How does your notion of flow-based forwarding > relate to**** > > the notions of L2 and L3 service in the problem statement and framework > drafts?**** > > ** ** > > We’ve had a prior discussion of IRB (combined L2/L3 implementation) and my > impression**** > > of the conclusion of that discussion is that the service is still L2, but > the L3**** > > implementation has been optimized. Is something similar going on here?*** > * > > ** ** > > Thanks, > --David**** > > ** ** > > *From:* [email protected] [mailto:[email protected]] *On Behalf > Of *Lizhong Jin > *Sent:* Thursday, July 25, 2013 11:52 AM > *To:* Thomas Narten > *Cc:* [email protected] > *Subject:* Re: [nvo3] Arch: Flow based forwarding service**** > > ** ** > > On Thu, Jul 25, 2013 at 12:53 AM, Thomas Narten <[email protected]> wrote: > **** > > > 1. Since TS is the end of the network, flow (e.g, 5 tuple) based VN > > forwarding would bring much network flexibility. The L2/L3 VN service is > > the traditional service, but I believe the VN should have the capability > to > > provide a flow based forwarding service. > > I'm not at all sure what it would mean for the VN itself (that is > NVO3) to provide "flow based forwarding service". What did you have in > mind? What benefits would that have?**** > > [Lizhong] flow based forwarding will not have influence to the underlay, > but the interface between NVE and NVA needs to exchange the flow entry, not > L2/L3 forwarding entry. There would be some scenario for flow based > forwarding, e.g, firewall needs to forward different flow to different VM. > The benefits is more finely control of the flow distribution.**** > > **** > > > In my mind, how traffic is forwarded by the underlay network is > entirely the business of the underlay. Flow based forwarding would be > fine, if desirable. But, that would be completely orthogonal to NVO3 > and the virtual network itself. I don't see right off the need for > NVO3 itself to be involved in this. > > Thomas**** > > ** ** >
_______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
