So what is up with SDN, and is it living up to the hype? Is the hype even still there like it used to be? I think the answer to these is that it is still out there. People are still working on it. And, kind of.
Using the Gartner Hype Cycle on Emerging Technologies we can see that SDN has pretty much topped the peak of inflated expectations and is headed into an area where things either persevere and survive or fall by the wayside.
As we enter this phase people are starting to figure out that the idea of your entire network becoming SDN focused, and that SDN meant that OpenFlow was going to be this driving force causing that to happen is just not very realistic in the near term. People have been excited about what SDN means to the network, and rightfully so, but the industry has struggled to find use cases that will support deploying it wholesale across your entire network. There are specific areas where it is extremely valuable and exposes powerful capabilities that you can leverage, but it is not the silver bullet to all your network problems.
Pure play SDN has been overshadowed by the emergence of NFV, which some argue is just specialized SDN…and to some extent they are right. NFV with no SDN component is not exactly something you should get too excited about implementing. However, this is an example of what I mentioned in the last paragraph. NFV is a specialized area where SDN is leveraged to do some pretty cool things like vCPE. Younger organizations, such as OPNFV have emerged that have sped right past the traditional SDN organizations like ONF with getting software releases out there that can used, at least in the lab, with no licensing costs. It even appears to me that the impetus for ONF starting to get traction with software, outside of OpenFlow itself, was the OPNFV and the momentum it has built around its open source solutions.
Another possibly telling thing that has happened is the departure from ONF of a key champion with Dan Pitt leaving ONF. Could it be that people are seeing that the vision of an OpenFlow focused SDN definition is just not what is going to happen in the end? I have long thought, and said in numerous blogs, that the idea of a programmable network is really what people are wanting. This does include the idea put forth by the rigid definition of SDN as being OpenFlow based network control, but also brings in technologies that bridge your path to SDN such as the Cisco/Tail-f NSO product as well as other OpenFlow-like control protocols. It brings in tools such as the Itential Pronghorn product, as well, that leverage those southbound systems to give you an automated operations capability that isn’t possible without them.
So, bottom line, in my opinion? The hype for SDN as stated in the recent past was more than ANYTHING could really live up to. We are entering a critical time where the tech behind it has to put up or shut up. We have to continue leveraging programmable networking techniques, in the meantime, to position ourselves for what is to come next…whether that is an OpenFlow based network control structure or something totally different. If you refer back to that diagram above from Gartner you’ll notice that the next phase will be Enlightenment. This is when maturity sets in, vendor solutions actually start doing what they say they will, open source solutions become production ready and usable, and our networks move to the next level. It WILL be programmable. It WILL be “software defined”. Will it be “SDN” as defined in the past? We’ll find out, but I don’t think so…