3 Horsemen of the Network Apocalypse: SDN

By Art Wittmann

**Editor’s Note: This is the second in a three-part series on the “netpocalypse.” Click here for part one, which discusses the evolution of virtualization.**

If there ever was a day in the ICT market where purely technical discussions drove the evolution of technology and standards, they’ve long since passed. The original idea of software-defined networking was that routing traffic based simply on a determination of the lowest-cost path ignored all sorts of interesting possibilities. Traffic could be routed on a certain path for reasons of security, for example, or because of the application that created it (video traffic goes this way, voice goes that way and data traffic goes a third way).

The other thing that SDN dreamers thought about was lowering the cost of the overall network. When each router has the brains (CPUs, memory, algorithms) to make its own autonomous routing decisions, that box is necessarily more expensive than a switch, whose forwarding tables are created for it by an outside controller.

Flexibility, economy and performance would seem to be good technical reasons to pursue SDN. Nothing political about that, right?

Well, perhaps. The networking business is Cisco’s business, particularly when you get past the busiest backbone segments, where Ciena and Juniper hold their own pretty well. Especially in the enterprise, it’s a Cisco world, with the company owning 65 percent to 70 percent of the switching and routing business and maintaining margins north of 60 percent. Cisco has managed to maintain those numbers for decades, much to the frustration of Juniper, Brocade, HP, Dell, Extreme and others. If only those vendors had a cudgel big enough to beat Cisco with, perhaps they could split the market a little more evenly.

Enter SDN.

Is That King Wearing Clothes?

Cisco has had three advantages that have allowed it to keep the majority of the networking market for itself. First, it has a sales organization second to none, including its indirect sales partners. When push comes to shove, Cisco finds ways to win and keep big deals. Second, the company has a support organization, including its partners, that is also second to none. Third, the company has great products – with proprietary software running on proprietary hardware that just works. The analogy I made here to Sun’s, HP’s or IBM’s proprietary RISC servers in the ’90s is stunning in its accuracy — except that Cisco has been more successful with its proprietary architecture for longer than the server folks.

One thing that keeps Cisco CEO John Chambers up at night …

… is the possibility of merchant-silicon-based systems displacing his high-margin offerings. I know this because Chambers has said so on more than one occasion. Chip makers, like Broadcom, play a key role in Chambers’ nightmare. It’s not too complicated to get some Broadcom parts and put together a Layer 3 switch with four 40-Gbps Ethernet uplinks and 24 or 48 10-Gbps ports for servers. There’s still some design work to do around things like ingress and egress buffer sizes, and of course developing a management software and an interface. Take a look at Arista’s products for a good example.

That need for a proprietary layer on top of merchant silicon has been a saving grace for Cisco. The thinking goes that at some point you’ll deal with proprietary interface, so you may as well stick with the one you know, even if you need to spend a bit more for the switch itself.

Enter SDN, which seeks to take the proprietary stuff out of the switch and put it into a software controller. Suddenly the game sounds a lot more like x86 servers that can fit into whatever virtualization, application and management scheme the end-user might pick. That’s why in 2012 VMware laid down a cool $1.2 billion for Nicira, and why Cisco in 2013 plunked down almost $900 million for Insieme Networks.

Cisco was caught fairly flat-footed by the interest in OpenFlow and SDN. While HP and others happily made some of their OpenFlow compatible so that research could go on, Cisco set to making some serious slideware. Just a month before VMware announced the Nicira acquisition, Cisco described its Open Network Environment to a rapt audience at its Live tradeshow. The move was reminiscent of Microsoft in the 1990s as it battled threats from Novell and the Linux community. Slideware is a good way for an incumbent to freeze a market.

The Cisco ONE vision in 2012 was exactly that. All vision and no product. Now, three years later, Cisco is talking about the products, but this time, it’s the prices that are missing. Cisco says those will come at the Cisco Live show later this year. The smart people at Cisco can interpret history as well as I can. They’re likely fully aware of the danger of overpricing the software components that will make up Cisco ONE, and the possibility of at least causing potential customers to evaluate other SDN options.

On the other hand, Wall Street couldn’t care less about Cisco’s dilemma. Hardware makers are out of favor and software makers are in, so Cisco already has a challenge to get investors excited. If it starts giving up its big, fat switching and routing gear margins, the Street will hammer the company hard.

Look for the final part of Art’s series on the netpocalypse on March 4.