Why Facebook's SDN switch won't affect Cisco's customers
- 24 June, 2014 05:29
Last week, Facebook announced a new product that's supposed to have the networking industry trembling. There were many news stories about Facebook's new homegrown SDN switch, known by the codename "Wedge" that's supposed to be the next big threat to Cisco and the traditional networking vendors. The operating system on the product runs Facebook's proprietary version of Linux called FBOSS.
Facebook plans to put the switch into the Open Compute Project, an open-source development community that Facebook has been active in. In many ways, Wedge isn't so much a switch as it is a reference design. Everything in the switch is open source -- software, processors, etc. - so anyone can build their own switch from the design. Now, for customers that are interested in this product, there's no "Facebook Store" to buy this from. Nor can organizations pick up the phone and call the friendly, neighborhood VAR. Instead, businesses would need to order it from a custom manufacturer.
So Facebook is now the latest vendor to toss its hat into the SDN ring. Are they a real threat to companies like Cisco, Brocade, and even the newly IPOed Arista Networks? While this product puts Facebook in the networking game on paper, I believe the appeal of Wedge will be niche and the impact negligible for a couple of reasons.
The first is that most mainstream businesses don't want "build it yourself, Open Compute Project" networking products. I've interviewed literally hundreds of customers and VARs on this topic and the appeal is very limited. Heck, most businesses don't even want to build their own servers and PCs, and those are about as commodity as IT infrastructure gets. This isn't Lego, it's a network, and "build it yourself" does carry with it significant risk. It's hard to overstate the value the network has today, particularly in light of the rise of cloud and mobile computing. The world has evolved to network-centric compute models, which means the quality, reliability, and resiliency of the network is the biggest determining factor of whether these initiatives are successful or not. The feedback I've gotten from most customers is that it's better to pay up and procure networking products from companies where performance is predictable and support is available. That's why the appeal of something like Wedge will be negligible in mainstream enterprises. Even if a company wants to test the white box concept, a product from someone like Pica8 that has a hardened OS and some technical support would make more sense.
Also, there's a channel issue. Most organizations buy networking products from their local VAR, systems integrator or other distributor. Facebook won't be available through these channels, so where would one even go to procure a Wedge? Most organizations don't have purchasing relationships with custom hardware vendors or contract manufacturers, and that creates another barrier to entry for Facebook and the other OCP products.
With that being said, I do think there is some appeal for a do-it-yourself switch for companies that need massive scale and custom features. Companies like Facebook, Baidu, Amazon, and Google have hundreds of network professionals to throw at projects like this. In these environments custom features and a product optimized for that environment can save millions. So when Facebook claims they've saved "over a billion" in hardware, I believe them. The operational costs may be through the roof, but I do see the appeal for these types of companies.
However, I also think if you're in this class of company, you're more likely to buy your own instead of running an "FBOSS" based switch. Why would Amazon or Google want to run a product running a Facebook OS? Would they even trust it?
I have no doubt that the hyper-scale companies are likely to prefer an open networking concept, I just think they'll create their own instead of using a reference design created by Facebook.
So, is there appeal for the Facebook Wedge? Sure, in Facebook's network. Other than that, let's stop counting Facebook's networking chickens before their hatched.