Software-defined networking is fundamentally about two things: the centralization of network intelligence to make smarter decisions, and the creation of a single (or smaller number of) administrative touch points to allow for streamlined operations and to promote workflow automation. The former can potentially lead to new capabilities that make networks better (or create new revenue streams), and the latter is about reducing the overall operating costs of managing a network.
Generating revenue makes perfect sense for the service providers who use their network primarily as a means to drive the business. But most enterprises use the network as an enabling entity, which means they are more interested in the bottom line than the top. For these network technology consumers, the notion of reducing costs can be extremely powerful.
But how do those OpEx savings manifest themselves?
OpEx you can measure
When we consider OpEx, it’s easy to point to the things that are measurable: space, power and cooling. So as enterprise customers examine various solutions, they will look at how many devices are required, and then how those devices consume space, power, and cooling. It is relatively straightforward to do these calculations and line up competing solutions. Essentially, you calculate the number of access ports and fix the oversubscription ratio. From there, you can do side-by-side comparisons in a mostly apples-to-apples comparison.
But SDN by itself does very little to impact the number or devices in a reference architecture. And when everyone is using roughly the same device components, these calculations start to converge on a narrow set of values.
It is worth noting that while SDN doesn’t change architectures a ton, there are other technologies that do. So I am not suggesting that all architectures are the same but rather that SDN doesn’t play the dominant role in determining what those architectures look like.
Beyond the easily measured
Once you get beyond the things that are easy to measure, the OpEx story gets a little bit tougher. To be fair on vendors, a lot of this is because most customer environments are poorly instrumented. It is difficult to know how much spend is associated with specific tasks or even certain parts of the network. Again, the default behavior is to lean on the things that are most easily measured.
At the top of that list is headcount. Most companies understand how human resources are distributed along different network boundaries. So the tendency is to look at a combination of the physical characteristics and headcount to arrive at a general number for OpEx.
So where do OpEx models fall down?
Imagine for a moment that you are a CIO (or if you are a CIO, simply reflect on reality a moment). As SDN solutions lay claim to reducing the number of people, consider how you are likely to respond over the first year of your newfound efficiency. If your new SDN-powered network does indeed require fewer people to manage, how likely are you to actually cut expenses and let people go?
If we are being honest, the answer is: exceedingly unlikely. Cutting headcount is one of those things that plays out well in slides and loosely constructed models, but operating a business is seldom that simple. More likely, your efficiency gains will be translated into a whole set of other things that your team can now do rather than deal with the mundane tasks of operating a network.
This is important though, because it means that the OpEx savings you based your decisions on might never really come to fruition—at least not in the way that you planned up front. And if you sold your decision based on these savings, you will find yourself in a difficult position during the next round of budget planning when you are forced to justify your existing staff in the face of promised reductions.
It takes a visionary
Just because more efficient operations didn’t lead you to cut half your staff doesn’t mean you did not achieve some benefit. The visionary who introduces a more manageable infrastructure will find that, over time, the operator-to-device ratio becomes more favorable over time. More simply, each operator can cover a larger number of devices, which means that you can grow your datacenter capacity faster than you staff new operators.
Even if you are not planning to add a ton of new capacity, the people that you have will be more effective. For example, if less time is spent battling network mechanics, more time can be spent doing the things that you know you ought to be doing but simply lack the time to do now—like documenting your operating procedures, or expanding your test capabilities, or even conducting more training and giving your team time to explore difference-making ideas.
The impact on SDN deployments
Promising OpEx is always tough, especially when it requires an initial capital outlay to get it. While the arguments can be compelling in the moment, there is just never a good time to spend money to save money. When times are tight, the dollars are hard to find. When things are booming, there are other things that are more pressing. And so the kinds of OpEx-friendly changes tend to linger on until some burning platform creates a compelling reason to leap from one architecture to another.
As vendors rely on some of these loose OpEx models, it means we could be in a position where the emergence of SDN is always right on the horizon. Like a mirage, it seems to move out a few more months every time we get closer. More to the point, we simply will not see mass SDN adoption purely for the sake of shaving off a few OpEx dollars. There has to be a compelling vision to get people to make the change…or maybe just a visionary.
The bottom line
It is extremely hard to put a value on doing those strategic things your staff never gets to do. How do you create a business case around developing your teams? How do you calculate in numbers the result of automation that reduces the number of human errors in your datacenter? How do explain that additional architectural review saves the business money? It’s not trivial. This is why it takes a visionary to cap investment in existing architectures and cross the chasm to embrace something new. There’s a reason that groundbreaking change is hard. The question for leaders is whether their job is simply to keep the lights on or to drive meaningful change into the business.
For the network engineers quietly fearing change or unnecessarily adopting a skeptical position, just know that the likely end game isn’t the elimination of your job. Rather you will likely be more critical than ever and less replaceable as you spend more of your time thinking about the hard problems as opposed to keying in simple configuration changes and executing excruciating troubleshooting procedures.
And for the vendors, it likely means looking more at opportunity costs than headcount costs as you construct your OpEx models. Of course this means knowing the customers you are selling to a bit better, which is probably a good thing for everyone.
[Today’s fun fact: Dolphins sleep with one eye open. It’s probably why they make good Navy SEALs]