In the hype-machine surrounding SDN and its role in complicated and intertwined data centers, enterprise networks,...
and service provider networks, there exists the surprisingly unrepresented market of small to medium-sized environments, which make up the majority of networks. They are too small for their own hyper-scale data centers but too large for an economical cloud option. They probably have a small team of IT folks, including just one person or a few specialists. They are your regional businesses; the larger employers that employ the majority of locals. Regardless, it's unlikely they are a huge enterprise or a hyper-scale cloud provider. A reasonable number of such companies likely don't have current plans for IT automation or to employ developers. How many network devices, such as switches, do most small businesses manage? For the vast majority, that number is likely under five. For these entities, is automation important? More realistically, is running an extra device, such as a controller that needs care and feeding -- user management, patch-level maintenance and resource monitoring -- causing more trouble than it is relieving?
Let's face it, SDN looks amazing. It looks like the answer to a number of issues depending on who you talk to and what day it is.
The value of a software-defined network is not realized, however, until the benefits of researching, studying, finding a use case for and then deploying and managing a new paradigm of network architecture and management far outweigh the operational and capital costs of doing so. Period. When is that? How many devices and high-caliber employees with particular skill sets are needed to successfully navigate such an endeavor? Realistically speaking, in today's day and age and with the maturity level of the technology that is labeled as "SDN," the number of devices is fairly high and the experience level of the engineer is reasonably deep. SDN really doesn't scale down. Looking at history, how many wireless access points represented the tipping point when a controller-based system became economical? On the high-end, manually tuning channels, power levels and other system variables starts to get hairy around six or seven wireless devices -- some would argue between three and five.
The difference is that with SDN, the technology is far more complex and, in many cases, involves new or proprietary protocols and technologies that still lack exposure to diverse environments. In addition, and probably more importantly, the skill set involved with running SDN today -- particularly OpenFlow -- is markedly different and arguably more extensive than a typical network engineer coming from a small to medium-sized environment will likely have, making SDN for small business complicated.
The development skills involved in running an SDN network are not to be underestimated. When deploying SDN for small business, not only will a given engineer or operator need to possess a deep understanding of an existing SDN environment -- including protocols, systems, architecture and monitoring -- but they will also be required to understand, in great depth, the details of the code that is talking to the network. SDN adds an abstraction layer that does not exist in traditional networks; it relocates the control plane.
The control plane is something many engineers currently take as a given. It exists as a foundation to view, change, connect, and disconnect systems, networks, devices, and entire environments. That layer now exists as its own external entity, residing in software that is, in many cases, written and supported by a third party. While large environments may have the resources to absorb and mold that kind of paradigm shift in their IT hierarchy, most small to medium-sized shops won't. And more importantly, they won't see a need to -- yet. Over time, as the technology matures, this gap will start to close, and the costs -- both operational and capital -- will drop. Just as with wireless networking, the number of devices that it makes sense to manage in a centralized manner will shrink, and the skill sets of engineers will either deepen to include more code or the products will become friendlier to non-developers. We've seen this pattern over and over in IT in the past -- it is not a question of if, but when.
In reality, small-scale SDN for small business may be worthwhile, just not yet.
What SDN enthusiasts can learn from Minecraft
Where does SDN stand in the large enterprise?
Some analysts say SDN gaining ground in mainstream enterprise