February 11, 2016 — When SDN first happened on the scene, there was a lot of expectation that it would be the panacea that would, seamlessly connect and control the network from a lofty perch in the control room. In theory, that is possible. SDN promised to centralize control of networks and make everything dynamic so all resources would be optimized, and available to wherever and whenever they are needed on the fly.
The theory is still sound, but the application is lacking across networks. And the reality of it is that implementing it is a lot harder than was thought at its inception. I am a fan of SDN and I think it will, once the bugs get ironed out, offer much of what has been hyped about it.
The problem with SDN isn’t its technology, it is that this technology, and the principles that go with it, do not work as expected in the field. There are many reasons for that, but the primary one is that SDN just won’t work with a single, centralized controller, as many have expected.
The networks are too diversified, with too many disparate platforms and legacy devices and too much proprietary hardware. In order for SDN to be fully realized, standards will need to be set and hardware will have to adhere to them. Period. There are other reasons. Vendors have yet to be all in on the platform, for one. So far, most have shown interest, and lip service, but few have invested in the development of it.
This has been an age old problem since the first chip hit the market. The solution has been, over the years, glue technology — something to connect the apples, oranges, strawberries, etc. But, glue technology in the SDN environment puts the brakes on speed and efficiency – what SDN is really designed to accomplish.
So until networks are designed and built with SDN in mind, and SDN gets a standard, it will just sit around for a while, with a few deployments. But it will be in good company because the Internet of Everything, and 5G are pretty much in the same boat.