Working for a small company, I sometimes have to put on my networking hat (which I haven’t worn consistently in a LOOOOONG time). We are blessed enough to have an AMAZING networking consultant, but if something goes awry while he’s working his day job, the troubleshooting and band-aids are primarily my responsibility. I also typically have a voice when new infrastructure is designed and hardware is purchased. These things, and my love of all things geek, have me keeping at least an eye in the FOSS Networking world.
Tomorrow evening RVaLUG (http://www.rvalug.org), my hometown Linux User Group, has the pleasure tomorrow of having Scott Clark, Senior Director at Vyatta (http://www.vyatta.com and http://www.vyatta.org), in town to talk about his company and how their products work. In the hope of having some intelligent questions to ask, I’ve been taking a look at Vyatta as well as pondering on some of the broader concepts they use.
The term “virtualized networking” confuses me a little bit. It seems that there are 2 concepts in indirect competition. One thought is that this means that you’re replacing datacenter switches and routers with virtual appliances. A second related but different thought is that you are securing an already virtualized network via standard networking principles in virtual appliances. #1 freaks out your local network admins. #2 is adding layers of encryption and additional networking stacks on a virtual network like Amazon’s AWS. Both present interesting challenges and different (?) solutions.
I’ve always assumed specialized networking equipment was, well…. specialized. I assumed those little blue boxes with all the flashing lights had specialized chipsets and algorithms running inside special chips that helped them move packets faster. Is that still the case? Or has a standard OTS (Off The Shelf) server today just surpassed the needs of 10Gbps? This article (from Vyatta’s CEO </full disclosure>) is talking about 20 Gbps Line Speeds using a single core Nehalem box. A 10Gbps for sub-$5k? Why yes, please. IF they’re as reliable as the $100k solutions out there now. Sadly, like email and telephones, networks have to “just work”.
To back this up even further, from http://www.vyatta.com/solutions:
The performance of off-the shelf x86 processors has increased over 100x in the past 4 years resulting in readily available systems capable of performing 10Gbps routing and security.
(Vyatta-specific) The “open core” model? Really? The open source version of VyattaOS is at best, limited. They want you to use it in your development and test environments, but in production you should move to their “Subscription Edition” (their words, not mine). I’m pretty happily on record at not being a huge fan of the “open core” business model, and I just don’t understand it here, either. One of the questions I’ll ask (and follow up on, of course, is how degrading their FOSS offering is a viable business model. Has anyone done it successfully yet?
Needless to say I’m looking forward to a great talk and discussion tomorrow. Thanks again to Vyatta and Scott Clark for taking the time out to come to Richmond.