Unless it's built at a lower layer than a piggy-back app, it's going to be a scale monster (IMO, if you plan for success, then ultimately you have highly connected NVEs - hello scale monster). Something like CFM for overlays (logical ethernet, logical continuity check built into your header as a special packet type, punt to momma NVE process ...not some cousin, etc?)? =8^) ________________________________________ From: [email protected] <[email protected]> on behalf of Dino Farinacci <[email protected]> Sent: Monday, November 18, 2013 12:09 PM To: Linda Dunbar Cc: Thomas Narten; [email protected]; Lucy Yong; Yves Hertoghs (yhertogh) Subject: Re: [nvo3] No need for NVE-NVE control plane [was Re: Poll for WG adoption and IPR check for draft-narten-nvo3-arch-01.txt
Dino, > >> If the destination NVEs don't >> exist, those packets go to black hole. >> >> Yes, it is. But a NVE could be unreachable and if the encapsulating NVE >> has a choice to select another NVE to encapsulate to, it should. This >> makes the architecture more robust. > > You don't need control plane protocol to achieve what you are asking here. A > simple TCP session among NVEs can do the job. When a NVE can't reach the > egress NVE to which the target is attached, drop the packet. That is a form of a control-plane protocol. And you will have to same scaling challenges with TCP and arguably it is worse, since TCP, inherently, does not have keepalives. Dino > > > Linda > _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3 _______________________________________________ nvo3 mailing list [email protected] https://www.ietf.org/mailman/listinfo/nvo3
