In the last post, I suggested that there is something more than commoditization driving SDN. Truth be told, I am not personally convinced commoditized networking solutions are a foregone conclusion. But the nuance here is that I used the word ‘solutions’ and not ‘hardware’.
Everyone and their dog has written about the ability of SDN to separate the forwarding plane and the control plane. The general idea here is that by extracting the smarts out of networking devices, the underlying hardware is responsible for just classifying flows and moving packets from point A to point B. In such a world, the ASICs can be merchant silicon. And if everyone is using the same merchant silicon, then the prices of the underlying hardware will drop. That’s right – no more 60+% margins on the hardware.
I suspect that this logic, when viewed from far enough way, holds water. If vendors converge on a small set of merchant silicon, the volumes will indeed go up on those ASICs, allowing guys like Broadcomm to drop prices even more. The end customers ought to reap at least some of that benefit. There are a few dynamics that could alter this a bit, like the emergence of Intel (potentially one of the biggest SDN winners), but that is a topic for another time.
Even with this idea in mind, I have a hard time tracking all the way to commoditized network solutions. It is clear to me that the underlying hardware prices ought to come down, but I don’t know that it necessarily means that the overall solutions reach commodity levels.
If you look at pricing today, in most networking gear, the hardware accounts for north of 80% of overall solution cost. In some platforms, it is actually 100% of the cost. If you take that mix of hardware and software pricing as fixed, it is easy to understand why commoditization is on the tip of everyone’s tongue. But is the pricing mix of hardware and software today going to hold?
Before I answer that, let me point out something. If you take any of the large networking vendors and look at their R&D spend, you will see that anywhere from 80 to 90 percent of that spend is on the software side of the house. But how can that be? I thought SDN was a few years away? Most of the value in today’s solutions is in the software. It has been that way for quite some time. The fact that the pricing doesn’t reflect that is an artifact of another day and time, and it’s a point that will most certainly change in the coming years.
So let’s assume that all the vendors converge on a small set of merchant silicon, which invariably leads to lower hardware prices. Even then, the network vendors will simply switch the pricing mix to more accurately reflect what we have known for years: that the software is where it’s at. While the hardware will indeed be cheaper, the overall solution will cost roughly the same. We don’t end up with commoditized network gear; we end up in a world that reflects the actual R&D splits a bit more faithfully.
Now I don’t mean to be all doom and gloom here. I do, in fact, think that even the solutions prices will come down. But it won’t be because of commoditization. One of the things SDN is doing is providing fertile ground for new ideas and new companies. This will breed a new level of competition that this industry has never really seen. And competition, my friends, is what will bring those nasty costs down. Now whether I think that costs are a function of capital expenditures or ongoing operational costs is a topic for a different day.