SDN is Dead! Long Live SDN!
And a Cautionary Tale
Feathers have been ruffled on several occasions over the last few years by someone publicly claiming that, “SDN is Dead!” This has inevitably been protested by leading communications service providers, rebutting that this cannot be true, because they are indeed deploying SDN solutions.
So like my mom used to say when I was squabbling with my sister, you’re right and she’s also right. If I expand the acronyms and add emphasis, it will become clear: “Software DEFINED Networking is dead, long live SOFTWARE defined networking.”
What’s the difference? I hear you ask.
Well, the first version says that SDN as a technology is dead, at least insofar as carrier networks. The goal of this technology was to commoditize network equipment, by gutting out their brains and differentiation, and have them all be slaves to a ‘master’ software controller that tells them precisely what to do. The pot at the end of this SDN rainbow was NE vendor independence, lowering the cost of the network, and an ability to craft differentiated solutions at the software control layer.
The defining element of this technology was supposed to be a set of standard protocols dictating precisely how NEs should behave. Well in fairness, this has partially occurred at the packet layer, primarily through YANG modeling for such items as how VRFs should be populated for creating VPNs, and there are attempts once again to create ROADM models at the optical layer. But seriously, SDN technology has not at all come into fruition as once envisioned, particularly for optical networking. Perhaps the best proof point of this is that OpenFlow optical activities are all but suspended.
So why didn’t this approach succeed. Primarily because the NE vendors weren’t having it. They agreed in principle, participated in industry forums, and then went ahead and did their own thing of delivering networking solutions with varying degrees of proprietary value. The CSPs needed to live with this because (1) they had always been reliant on the vendor community for technology evolution, and (2) the vendors, not being totally clueless, started opening up their solutions to software control, but on their terms. So things worked as before, with shifts in orientation.
Which brings us to the second point, SDN as a business goal is in fact being realized. This goal is speeding up service delivery and maximizing use of network resources. In fact, I would argue that this goal can be summed up in a single word – AUTOMATION – and that it is being achieved through the use of SOFTWARE.
What in fact is happening is that the industry is relying on all manner of semi-standard and proprietary interfaces from NE vendors (often consolidated by SDN controllers or next gen NMS), and focusing huge amounts of resources on the software layer to achieve service agility and operational efficiency through automation. We see this in a slew of new industry initiatives like ONAP and OpenNFV. It is also evidenced by many industry conferences that used to be wholly focused on networking equipment and are now trying to stay relevant by refocusing on topics like orchestration, cloud-native software, intent-based controls – all based on or related to automation – and now starting to be accompanied by AI and machine learning.
The value chain is being turned upside down. In the past the network delivered the value and this was supported by network management systems. In the new world, the network enables value (something still needs to perform the heavy lifting of carrying bits from A to B), but the value itself is coming from the automation software layer that controls the network.
I was discussing this with an industry analyst friend, who remarked that this can be seen as the network management folks (after many years of playing a lowly supporting role) coming into a cycle of dominance. I disagreed. My own conclusion is that it is the power is shifting into the hands of the DevOps and Information Technology (IT) folks who know how to structure and extract the most value from the cloud native software applications, which will power this automation.
There is a cautionary tale here. While we are achieving the business goals of SDN in the near-term, we are also paving the way for the swallowing of our telecom industry in the longer term by those Internet giants that have already steamrollered over other industry sectors.