There’s a lot of buzz around software defined networking (SDN) and even more head spinning mumbo jumbo mixed in with the hype. So, I’m guessing that many folks are asking – Why should I care?
Pardon the presumption, but I’ll take a shot at explaining it.
The answer starts with the Cloud (oh, here we go again). The Cloud is all about economics. Companies of all sizes want to run applications out of the cloud because it costs less and involves less hands-on maintenance to keep it running. There are some downsides, like creating new silos; losing visibility into some business process; and relinquishing some control over the health and accessibility of some key business functions. But, overall, the Cloud is about doing more while spending less.
One of the key concepts in the Cloud world is what you might call “on-demand virtualization” or just “virtualization.” Basically, the cloud shares it own infrastructure by spooling up and shutting down instances of applications, running on virtual servers, on an as-needed basis. Conceptually, this is similar to how networks shifted from synchronous to asynchronous forms of multiplexing. The bottom line is, when you optimize how you’re using available resources, you don’t need as much gear to operate. In the cloud world, this has been pretty effective thus far. Sure, there are plenty of instances where demand has knocked the cloud-model down, but cloud is still winning on its efficiencies and economics.
SDN is becoming more interesting in part because networking hasn’t really kept up with the Cloud. Applications running “in the Cloud” need to be accessible via networks, and increasingly need to talk to each other by using networks. If you want to break down silos and integrate across clouds; or solve the cloud-to-cloud problem as we’ve dubbed it (or the multi-cloud problem, as TM Forum calls it); or even just have applications talk to each other in the same “cloud” (i.e. data center) then you need networks to function in ways that mesh with virtualization and cloud economics. Right now, that isn’t the case.
The problem techies are running into is that they still have to do a large amount of manual or semi-automated network configuration in order to get applications to talk to each other. There’s very little capability within or across clouds – or data centers – to have the right kinds of network connectivity pop up and down on an as-needed basis in sync with the on-demand virtualization process. What I mean by “right kind” of networking means networking with security, QoS, SLAs, and other configurable factors that are appropriate to the applications that use them and the transactions they facilitate. So what happens? You may have to nail up, and over-provision, the underlying networks in order to facilitate what cloud-based applications are trying to do. When that happens, you’re no longer optimizing your resources. The more complex your cloud gets, the uglier, more expensive, and potentially more risky the networking components of your cloud-based applications and processes become.
The point of SDN is to fix this problem and make networks work in smarter ways that play nicely with how cloud-based applications want to work. There are other applications for SDNs, but right now this seems like it would be the biggest driver of SDN adoption given how much is moving to the Cloud.
So, what’s the problem with SDNs? Well, simply to play devil’s advocate, I’d say the problem is probably software. A highly respected CIO once told me that the beauty of hardware was that it’s scientific – if the light is green, it’s working. If it’s red, you plug in another box. Software, he said, is more like black magic. So, that could be the trouble with SDNs. They introduce the variability of software. So, while they may add more intelligence, control, and flexibility to cloud-enabling networks, they may also introduce more implementation risks and maintenance burdens, as software historically has done.
SDN’s evolution will be interesting to watch, but it would probably be another of those under-the-covers networking ideas hidden in a closet somewhere if it wasn’t for all of the attention currently focused on the Cloud.