There was an interesting Twitter thread over the weekend initiated by Ethan Banks (@ecbanks). He commented that there was too much technique churn in SDN and NetOps (the networking equivalent of DevOps). His point was that in the face of all the change in how to do things, it left users in an impossible spot. How can up pick up a new technology if the frameworks around how to use it are consistently changing?
His conclusion was that we cannot herd these cats. But what is really going on?
No consensus on operating models
The most basic truth here is that there is no real consensus on operating models around any of the new technology. While there are rough agreements on a few architectural principles (and even there, far more is in the air than well grounded), there is really not a lot of best practices to which companies can pin their operations.
Sure, it might be obvious to people that SDN is here to stay. But what exactly does that mean? And which SDN do I evaluate, purchase, and eventually deploy? Do I go with OpenFlow because ONF has convinced me that openness is the primary tenet? Do I lean towards overlays because I believe VMWare when they talk about edge policy management? Maybe I go with ACI because the thought of a tightly-integrated infrastructure centered on the applications sounds good? Maybe I check out one of the startups that has various SDN implementations.
There just isn’t a common deployment model. So trying to do anything general becomes hard.
The Great Migration
Now, not having best practices isn’t always an inhibitor, but out industry moves with the grace and speed of a newborn horse.
For those of you who have never been to Africa, you need to go. We went for our honeymoon. It was September, and we caught the tail end of the Great Migration, where all the zebra and wildebeest migrate north to the water. It is sometimes called the Greatest Spectacle on Earth. It’s a long line of millions of animals heading the same direction.
The real action is when they get to a water crossing. It’s typically a small drop (maybe a few feet) and then a river. The crocodiles know the animals are coming, so they camp out below. And then all the zebra and wildebeest group up at the edge, none of them willing to make the leap. At some point, one of them goes. I don’t know if he just decides, or gets pushed, or maybe he just loses his balance and goes. Whatever the reason, once he goes, they all go.
As we look at IT generally and networking specifically, we have the same dynamic. No one wants to be first, because they know there is danger in trying to make the crossing. Of course, everyone knows there is a place to be that is better than where they are today. And so we all just wait until someone makes the move. Once they make the move, then everyone follows in exactly the same way. This is why the crossing happens at one point. It would seem obvious that animals who were smart would walk a few feet upstream and jump there, but there is no proof that you can make it there. So they cross where the crocs are.
Operating models and the migration
The challenge in not reaching consensus on an operating model is that the people who are looking to follow a set path are unable to find one. You want to know why deployments are slow? This is why. All the buzz about this new thing or that new take on something actually slows everyone down.
But the real dynamic here is not that people talk about different things. It’s a nascent space, so that’s expected (maybe even encouraged). The real problem here is that the people making most of the noise represent a tiny, tiny, tiny fraction of the people actually buying stuff. When you get into the echo chambers created in the Twitterverse and the Blogosphere, the people who are vocal and advocating positions are a very small handful who have an extremely heavy hand in shaping the dialogue across the industry.
The challenge here is that if you listen to the vocal minority, you are drawn to the next cool thing. And the next. And the next. Before ever having moved on to the first thing. The pursuit of coolness is fun. It’s validating. Thinking up new stuff is exciting. But it doesn’t solve common everyday problems, and it flies in the face of how people move in this industry. We have people pooling up along the side of a river, and the leaders keep running around finding new places to cross the river without ever really crossing it. Is it any wonder that deployments are lagging?
The risk of following the few
The risk here is tremendous. If everyone is waiting to cross and they cannot find a place to cross, at some point they all just go home. It behooves us as an industry to start to narrow down the choices some and develop some best practices. This is why open source efforts like OpenDaylight are interesting. Yes, it’s open source, but more importantly, it’s a way to get people to agree on where and how to cross the river.
But in the absence of having a large architectural framework to work around, the other key is to demonstrate that the river can indeed be crossed, even if only to a subset of the masses. Very specific, very controlled solutions will garner attention. They might not be the most full-blown things. They might not be the sexiest solutions. They might not even have all the buzzwords. But if they solve an actual problem, they will be deployed. And then people will follow.
The bottom line
I used to think that the large companies couldn’t get it right. But they have been fooling the market all along. With so much churn, they continue to sell a ton of kit to people who cannot make the leap across the river. While competition keeps picking new crossing points, the big guys just sit back and make a killing. They aren’t stalling SDN; they are just being pragmatic. It’s not that they are incapable or even unwilling to do something new. They just don’t need to if their objective is to keep making money.
Of course, this leaves them exposed to someone who does deliver a solution to an actual problem. Because as they go one, so go they all. And it might not be the leading players who make up the vocal minority. It’s the mid market and lower fringes of the big markets that will determine the winners and losers. The question for vendors is what do they want? I suspect it isn’t another shiny knob. They want operational sanity. Deliver that and you will see the Greatest Spectacle on Earth play out in our very own industry.
[Today’s fun fact: A one-day weather forecast requires 10 billion math calculations. Now that’s some Big Data. Imagine completing all that and still getting it wrong!]