Recently Cisco’s annual user conference, Cisco Live!, was held in Orlando, Florida. This is the place one wants to go to learn about networking trends as evidenced by the 20,000 attendees. One of the hottest topics there was “software defined networks” (SDNs), as network professionals try to figure out what this trend means for their careers and the choice of network infrastructure.
Much of the media focus on SDNs has been on how they will commoditize network infrastructure, since organizations will be able to build networks using a mix of low-cost switches from any number different white-box switch vendors, combined with a controller from one of many start ups. This ‘mix & match’ model is fine for companies like Amazon, Baidu, or Facebook, since these organizations have the skill set and the personnel to do things like write their own operating systems and create custom functionality in the network. In a sense these large organizations trade off infrastructure complexity with people complexity.
So, are the strategies of the Googles (and other giants) of the world a sign of things to come? My answer is an emphatic “No.” They have been, and always will be, anomalies. Their way is a little like having a NASCAR crew — a group of people who have the skill set to take nothing but parts and make a car which performs better than anything than the average person can buy from a dealer. Most of us, though, just don’t have that skill set — we buy a car that’s prebuilt because we want a simple driving experience without having to go through the complexity of building a car.
I believe what network managers want is something along the lines of my car analogy — a network that is easy to manage, not simply something that’s cheap to buy. Think about the bang for the buck here: of all IT spending, hardware (all hardware, not just networking) accounts for about 17% of spend. Within that, networking hardware is about 20% of that 17% meaning it’s under 4% of all IT spend. Lets say someone could knock 25% off the total spend on network hardware, that would mean a 1% total savings on IT spending. Whoop dee doo. Which brings us back to the Web giants — these companies don’t buy white-box hardware to save money, they do what they do to create unique functionality.
For the rest of the companies out there that want a network that’s simple to manage, I believe they’ll gravitate to companies like Cisco, Brocade, and other, more traditional network vendors that can deliver end-to-end solutions. Why? Because the fact is that building something that’s simple to use actually takes a ton of work behind the scenes. Back to the car example — my wife’s Mercedes has a bunch of functionality that makes the driving experience simple. Over time it learns her voice so the speech recognition works great, and she can hit one button and the seats reconfigure to her. Why did we pay up for this kind of functionality? Well, because she told me to… but also, she wanted a high-quality driving experience that’s easy to use.
For most mainstream network managers, the value proposition of SDNs is to deploy a network that’s easier to manage than networks deployed today. I believe that means more end-to-end functionality within a single vendor solution, not less. The big winners in the SDN market will be the vendors that are able to create a simplified front end by doing the hard work in the back end. So let’s put an end to the talk that simplification of the network leads to commoditization. Instead we should understand that building things that are simple to use is actually very hard.