There is probably never going to be a perfect balance in the industry between Do-it-yourself (DIY) and Do-it-for-you (DIFY) networking. It seems exceedingly unlikely that there is a one-size-fits-all type of solution out there. And so we will invariably end up with a bifurcated market that requires multiple solutions for its constituents. But if there is not a perfect balance, which one of these is likely to see the most action?
If you were to base your guess on industry chatter, you would have to conclude that DIY has the upper hand.
There is a ton of momentum right now with both SDN and bare metal switching. On the SDN front, it is all about orchestration and automation. The ability to streamline customized workflows is attractive, especially for the large IT shops that sink tens of millions of dollars into managing their monstrosities. Once you get into anything that is customized, there is a degree of DIY-ness that is required. No product is designed expressly for your particular environment, so you need to the ability to customize what you buy to do what you want. Beyond that, there is an awful lot of talk about APIs and programmability.
Bare metal switching is a different initiative with different objectives that end up in a similar DIY framework. The move towards a more server-like environment allows users to customize their switching solution. There is great power in having absolute control over how a device behaves. It allows users to pick and choose tools they are already familiar with, extending their functionality into the networking realm.
However, the challenge in using industry dialogue to conclude where things will end up is that the chatter does not always match exactly the buying patterns. Indeed, public discourse most typically leads broad deployment – sometimes by several years or more (think IPv6, Internet of Things, or even electric vehicles).
The DIY movement in networking is real, but what is it about? The ability to tailor specific networking applications to the infrastructure is about eking out performance or customizing experience. It is about modifying a base set of functionality to fit better into your specific context.
For this to matter, you have to be pushing the envelope in terms of performance or capability. But the truth is that the bulk of the networking space is simply not there. Their issues are not in customization. They want to be spending less time with the network, not more. The problem they need solved is more about operating their infrastructure and less about creating substrates to connect it all together in some unique configuration.
But you don’t hear from these people in industry forums and on social media. They lack the interest, time, and sometimes confidence to express a point of view that is less visionary and more functional. As a result, we only hear one side of the story. It plays out in blogs, on Twitter, in press articles, and on conferences stages. And with every word and unapproachable idea, we collectively push the majority of users further into the background.
The solution here isn’t to retreat from change. But we need to make sure that new technology is usable for the legions of people for whom the network is primarily a means of enabling their business. We need to advance with equal enthusiasm DIFY networking.
So why don’t we do this naturally as an industry?
There are two major dynamics at play. First, incumbents tend to be capability-driven. Customer X needs something, so they build whatever widget is required. The focus is on the capability, not necessarily on how that capability is inserted into a widely consumable workflow. If there is any doubt here, ask yourself if networking workflows today are more arcane or intuitive. And then ask yourself why certifications are so important. The only way to validate that you have mastered the arcane is to produce your certificate as proof.
The second dynamic is that new initiatives (be they new companies or just new projects) tend to target the hot spots. Those hot spots are identified by the vocal minority. And networking’s vocal bunch consists of strong proponents for customization, primarily through tooling and development frameworks.
But even here, customization is rarely the outright goal. Unless your business requires differentiated network services (a la service or cloud providers), you likely don’t want to be customized for the sake of being customized. Rather, the customization trends are a response to a broad deficiency in the networking industry. More directly, if my vendor cannot give me what I need, at least give me the tools so I can do it myself.
Both SDN and white box switching are great movements, but they are responses to a long-time issue with legacy networks: the equipment is needlessly expensive, and networks are ridiculously hard to manage. When these issues go unaddressed for decades, what are customers to do? They stand up and collectively say “Screw it. I’ll do it myself.”
When the DIY trends exist long enough, we end up fooling ourselves into thinking customization is the goal when all along it was merely a workaround. We replace intuitive networking with There’s an API for that networking. Essentially, we have shifted the cost from procurement (you can buy cheaper equipment) to development (but you have to customize everything around it).
This doesn’t seem right. I suspect the right outcome for the industry is to take the technological advances, develop them to completeness, and deliver an infrastructure that delivers. Such an infrastructure could still have the hooks for the DIYers, but it would be functional for the DIFYers as well.[Today’s fun fact: If all of the oceans in the world evaporated, Hawaii would be the tallest mountain in the world. Take that, Everest!]