These new technologies are both impressive and useful in several ways, but I doubt the high-end switch vendors will be losing any sleep over them yet.
People have been talking about SDN and running Linux-based white box switches that cost significantly less than the name brand gear for years now. So far, reviews of actual deployments have been mixed, and the market reaction has been cautious.
Even those platforms are still ultimately running custom ASICs for the actual switching functions, which gives them a natural and very significant advantage over anything running under Linux on a general purpose CPU, so it's hard to see the likes of Snabb disrupting the same market any time soon.
The real value in these new technologies is the extra flexibility they offer precisely because on a general purpose CPU you can do whatever you want with the traffic as long as it fits in your time budget. That opens the door to a level of customisation in network functions that has been hard to achieve in the past, but right now I'd probably be more worried if I sold specialist networking devices with five/six figure price tags than if I sold standard, high-bandwidth switches and routers to the same sort of customer.
> Won't be too long before there are nice ecosystems built around DPDK, Snabb, etc...and people start replacing their higher end Cisco gear.
actually, might it possibly go the same way as it happened earlier in graphical workstation market ? where need lower end gets squeezed out by these commodity off-the-shelf components, and higher end stuff slowly dwindle into irrelevance...
The similarities are there, and they start getting nervous. They are not only threatened by open source running on x86 hardware, but also by open source running on commodity switches. x86 Hardware achieves impressive speeds but consumes to much power compared with special purpose chip sets. If only open-source friendly Intel had switch chips comparable with Broadcoms' (which is notoriously hostile to free software)..
Lisp people might recognise the name of one of the main Snabb authors, Luke Gorrie. He started Slime. Always fun to see what he's up to. Andy Wingo, the presenter of the talk, works on Guile.
also was involved in OLPC and distel (https://github.com/massemanet/distel) i.e. distributed+concurrent emacs lisp (primarily used for erlang development within emacs)
SNABB is not just a great networking tool, its also one of the best users of the Lua VM out there .. I've learned so much for my Lua-based job, just by looking through the SNABB sources. I highly recommend anyone getting into Lua to have a few days worth of code-reading ..
Very exciting topic - I would be really interested to see how various features such as Link aggregation (LACP) and Spanning tree protocol would be implemented within the framework.
I remember the various RISC UNIX vendors ignoring Linux when it was new, a bit clunky, and short on enterprise features. They missed the head start.
Won't be too long before there are nice ecosystems built around DPDK, Snabb, etc...and people start replacing their higher end Cisco gear.