Hi, On 08 Feb 2016, at 16:24 , Jean-Yves Migeon <j...@netbsd.org> wrote:
> Dear list, > > I know that this question is not NetBSD specific per see, but I am aware that > some of you out there have implemented and used vxlan-like technologies to > extend networks at L2. Some have worked on it with rump [1]. > > $DAYJOB is currently investigating using these kind of technologies (mostly: > vxlan or otv) to extent "seamlessly" L2 networks between two "zones", but as > I have never used such protocols I am bit at a loss here, especially on the > administrativia stuff. > > If any of you used or use it and is willing to share (privately or not) his > opinion on these, I will gladly take it, especially for things at a larger > scale than what I could achieve on my small test bench. I am bit wary of L2 > extension protocols when it regards scalability of the infrastructure, > specifically for large networks (I guess that an alternative to STP needs to > be used here?). Then you say “implemented", are you referring to “coded an implementation of VXLAN” or “installed something that knows VXLAN and put it in production” ? I’ve done the latter, and it works very well. It solved the same problem your $DAYJOB wants to solve for us. Instead of trying to stretch a LAG in-between the different locations (the old solution, we now use a VXLAN tunnel in-between “leaf" switches at the the different locations. We use eBGP in-between “leaf" <> “spine” switches/routers. Re, /P
