I think we’ve been here before…

I’ve been failing miserably at getting some posts written. Working with VxLAN and leaf/spine design a lot lately, and also trying to get 1kv working on KVM (serious dearth of documentation there) though, so hopefully there will be some cool new stuff soon.

In the mean time, I just figured I’d write a short post about how crazy and cyclical this industry is in general. As I mentioned I’ve been living the VxLAN dream lately which is a blast, but I can’t help but to notice how silly VxLAN is at the end of the day. Don’t get me wrong I think its super cool. The scalability and flexibility that it affords is absolutely awesome. I also happen to believe that its ‘won’ the overlay war and that we will be stuck with it (for better or worse) for at least a little while.

As I’ve been doing some more lab work with it this weekend though I was lamenting that I don’t have a Cisco Nexus 1110X appliance. That appliance is basically just a C-series UCS server, but it has some magical properties that allow ‘blades’ to be installed and essentially associated to a 1kv VSM. One of those blades, and the one I’m yearning after, is the VxLAN Gateway blade. As I was being a sad panda at not having this cool tool, and thinking about how not cool it is that my VxLAN lab VMs are stuck in overlay land without any access to the rest of my network, I had a thought… Having just been playing with bridges in Ubuntu (while monkeying with KVM) I thought to myself that the VxLAN gateway is really just a magical bridge that binds VxLANs to VLANs. I figured if thats all it is there’s no reason I can’t do that with a router and BVIs. BVIs are cool… I guess….

So I deployed a new CSR 1000v to test this out, and I got very excited as “bridge irb” was accepted syntax, but alas when going to my interfaces ‘b?’ failed me and there was no bridge-group command. Down, but not out I tried to download a demo of Vyatta (Brocade’s virtual router) and promptly gave up as it wouldn’t let me change any interfaces (barking about licensing even though I was on the free trial guy — weird). I decided at this point to drink a beer…

Feeling much refreshed I decided that I didn’t need no stinkin’ router anyway and that I would just do it in Linux. 10 minutes later, my test bridge-domain was pulling DHCP from my little 881 and humming right along. One ESX host in subnet A (with the 881, and the VLAN it was being bridge to) and a second in subnet B. BOOM! Layer 2 adjacency across an OSPF backbone via my friend VxLAN, and magical bridgie-ness via Ubuntu.

So I said that this was about how cyclic things are in networking. If you hadn’t caught on yet the most obvious piece is that VxLAN, while new and shiny, is really just a glorified VLAN. VLANs in case you hadn’t noticed have been around for a LONG time. In fact a quick check on wikipedia (so this is obviously an indisputable fact…) VLANs came about somewhere in 1988!! Thats 26 years ago. Twenty-six years and all we’ve managed to do is to add a single letter to the name!? Of course its cooler than that, but at the end of the day VxLAN really is just another layer 2 segmentation technology.

Thats not all though. Bridging… whoa thats old too! I love talking to guys that sat the CCIE R/S way back when (two-day lab) and listening to them talk about bridging IPX and Appletalk! Another quick wiki search shows an RFC (1286) from 1991! My duct tape and bailing wire VxLAN gateway, while a bit (lot) kludgy, is not much different from a software or hardware bridge (it’s just not actually participating as a VTEP like a real gateway would), and at the end of the day it’s just a fancy bridge.

Those are just the direct comparisons of technologies; more parallels can be drawn in the never-ending battle of centralization vs decentralization. At any rate, I think its pretty funny that in 20+ years we are still doing basically the same things, just with cooler sounding acronyms. I suspect that there is a lot we can learn from the history and evolution of both VLANs and bridging that will be directly applicable to VxLAN. It’s probably best that we do some learning of our history of networking so that we don’t make all the same mistakes of the past twenty years all over again!

Advertisements