SDN is coming and it’s going to be great. The enterprise will finally unshackle itself from its network hardware chains, and a brave new world of scalable, malleable and eminently less expensive data architectures will be at hand.
All of this is undoubtedly true, but I can’t help thinking that this is exactly what I heard as a young man back in the early 1980s, when PCs were suddenly all the rage. A home computer is vital, we were told, because the whole world was going digital and you don’t want to be left behind. Besides, it can take care of budgeting and appointments, help with homework (as if) and even prepare the weekly grocery list. After about two months, the vast majority of home PCs sat in the corner of the living room collecting dust.
The thing is, none of the claims that early PC boosters made were false. It’s just that it took more than a collection of isolated hardware to fulfill the promises. The key was networking. Once the home PC could connect to millions other PCs, then financing, shopping and, yes, even homework, was made easier.
So when it comes to SDN, I have to wonder if we could just be repeating the same pattern on the enterprise scale. My central question is, and always has been, what exactly can you do with all of this network abstraction, and is it truly the value proposition that supporters make it out to be?
According to KPMG’s Adrian Bridgewater, SDN is simply the means to an end, and that end just happens to be a completely different data stack than exists today. With abstraction comes the ability to isolate discrete digital constructs, such as containers and microservices. Not only is this a boon to security, but it also allows organizations to decentralize their data architectures across the cloud to enable rapid and dynamic resource configuration and deployment. And the best part is that this can all take place on the application layer, no need for an overarching network management platform at all.
This will be crucial as increasingly disruptive (to traditional networks anyway) changes emerge in the digital economy, says CIO’s Bonnie Gardiner. In case you haven’t noticed, innovations like the cloud, data mobility, Big Data and the Internet of Things are arriving at an increasingly rapid pace, and the last thing the enterprise needs is an inflexible network infrastructure based on technologies developed 30 years ago. If nothing else, SDN allows the enterprise to build and deploy applications and resources based on what the business model requires rather than what the network can support.
SDN isn’t about networking as much as it is about network control, according to a recent white paper from Allied Telesis. The main flaw in current architectures is that individual nodes do not have a broader view of the network ecosystem, so they simply forward packets in ways that seem logical from their limited perspective. With a centralized control plane, forwarding decisions can be made based on input from multiple nodes, producing more streamlined traffic flows. In this way, SDN will bring three key elements to the network stack: increased intelligence on the node, enhanced analytics on the controller, and network-aware applications that can navigate this environment on their own.
This points up one of the central truths about SDN: changes to the network alone cannot produce all of the benefits that people have envisioned, just like the early PC. Virtualization across the entire data ecosystem, including the data and applications, will be required to truly bring about the dynamism, scalability and self-management that knowledge workers are demanding. And ultimately this will shift the way we look at the full spectrum of our data operations, from the way workflows are managed to the type of business processes that are supported.
This is the primary reason why the enterprise should avoid getting too caught up in the SDN hype. Sure, it can lead to a brave new world of computing, but only if the entire data stack is pushed forward in a complimentary manner. In the meantime, organizations still have to puzzle out how exactly this grand, integrated data ecosystem can best be leveraged to support broader business goals.