A few months ago, I noticed that the networking cognoscenti were becoming jaded about software-defined networking (SDN). To be fair, the networking cognoscenti can skew toward disgruntlement, so it was no surprise to see this restive bunch cast a jaundiced eye toward networking’s greatest, latest hope.
I consider myself among the skeptical and wary, always cognizant that vendors can be inclined to advance a self-serving agenda that sometimes is designed to satisfy their own near-term interests over the long-term objectives of their customers. That works particularly well when the vendors can trick the customers into believing that they’re actually looking out for them. As our ancient forebears knew, caveat emptor was more than a catchphrase.
Asking Why
All of which brings me to a puzzling aspect of the current disaffection with SDN, expressed most recently in a highly readable and strongly recommended post by Ethan Banks of PacketPushers fame. My question, which I put to Banks to and to everyone else for whom SDN has become an annoyance, is simple: Are you really upset with SDN, or are you actually frustrated with the way the term has been used and abused by the vendor community?
It’s not an academic or an idle question.
One should remember that SDN, properly defined and understood, is a creation of a customer-centric consortium, the Open Networking Foundation (ONF), not a marketing or technical construct espoused by a given networking vendor or even by a group of vendors. If the term “SDN” is being bastardized and demeaned, it is not the ONF that is doing it. More directly, if the term is being cheapened, the devaluation is occurring at the hands of vendors.
But why? There are at least two possibilities. One is that certain networking vendors want to exploit the positive connotations, the afterglow, that surrounded software-defined networking (SDN). According to this theory, the damage they’re inflicting to the SDN brand is unintentional and ironic: They wanted to ride SDN’s relatively pristine coattails, not pull it into a seedy gutter of disrepute. I would be inclined to accept this theory if vendors adopted SDN definitions that accorded with that of the ONF, but. for the most part, that’s not what’s happened.
Agents Provocateurs: Back in Action
Instead, vendors typically recast SDN in forms that correspond with product roadmaps and company-specific strategic objectives. The result has been market confusion and cynicism, understandably so. When a term is spun to mean practically anything to anyone, it risks losing its specificity and its relevance.
Allow me suggest that at least a few vendors would be neither inconvenienced nor unduly troubled to see SDN’s identity fractured and splintered like a broken mirror. It would not be the first time that fear, uncertainty, and doubt were deployed as agents provocateurs in a commercial context.
Nonetheless, coming back to my question above, I would counsel that we think carefully about whether our annoyance is really with SDN or with the way the term “SDN” is being manipulated and distorted by the vendor community.
As always, it is helpful to diagnose not only what is happening, but to try to understand why it is happening, too.
Twilight in the Valley of the Nerds 
Would love your take on Cisco apparently ceasing sale of ACE load balancers and shifting development resources elsewhere. Does this mean Cisco is exiting the Layer 4-7 market or just shifting emphasis to virtual solutions on their Unified Networking Services platform.
Great question, and well timed. I intend to post on that topic tomorrow.
Perfectly said Brad, it sure will be fun to look back on this all someday. We will look like cavemen.
I think the annoyance is clearly with the way “SDN” is being manipulated.
If we start with the ONF whitepaper (at http://www.opennetworking.org) that defines the SDN architecture and other things, we have atleast three camps today –
1. those following the ONF definition
2. those who think the infrastructure and control layer can be kept combined just as today and only build a new application layer (compared to the SDN architecture)
3. those “tagging on” to the coattails (you’re being very nice with your comments above 🙂
I personally think the second camp shouldn’t tag SDN to their solution as it’s a different architecture where the infrastructure and control layers are one but distributed across each network device. Something else like ACN (Application Controlled Networking) may be more appropriate.
We’re left with who polices the third camp. We need someone as a watchdog and fact-checker, but ONF is not stepping up to the plate and nor is the media which is actually helping to sensationalize the announcements from this camp.
Brent,
I respect your knowledge and what you’ve brought to the community through your thought leadership, but I have to disagree that the ONF is innocent when it comes to creating confusion around what SDN is. SDN is so not about openflow that it is almost becoming comical. OpenFlow is a wire protocol for configuring flow tables and did a great job of generalizing and abstracting that concept. SDN can easily include that among the many abstractions that need to be defined to get to the software defined datacenter nirvana. It is completely unfortunate that since it’s initial press release announcing it’s formation, the ONF has encouraged the huge number of people I’ve met who think that acronym stands for OpenFlow Forum (clearly they have bad phonetic skills too). The OpenFlow tail was allowed to wag the SDN dog long enough that wholesale SDN-washing has been unleashed. The ONF needs to figure out how to re-position to stay relevant. It’s truly unfortunate that this wasn’t founded as the SDN Forum or similar, which would fit nicely around standardizing northbound APIs also, These need to include distributed-application and network abstractions to go above the device abstractions baked into OpenFlow and some OF controllers of note. APIs for machine to machine conversations across the layers of the new datacenter are where the action is commercially, intellectually and increasingly. And by the way, the big vendors have done a pretty good job of insinuating themselves into the usual standards bodies leadership roles at the ONF that have given them unlimited filibuster in the IETF and similar standards bodies. The board is “customers”, but I think the vendors have gotten under the wire in the working groups.
I am still hopeful however, that the average IT thought follower is only about 9 months behind people like you in terms of seeing what’s real through the fog and being able to see how it applies in the datacenter. Past all of the hype cycles of the various buzzwords, we are talking about an opportunity to re-factor our network model and make the solution for e.g. multi-tenant IAAS “as simple as possible, but not simpler” (nobody will even try to argue with Einstein). This is in contrast to the autonomous systems beast that IT is currently trying unsuccessfully to tame at webscale.
Great work Brent.
Pingback: Arista hires a CFO, poaches a big shot Cisco executive ← techtings
Pingback: Arista hires a CFO, poaches a big shot Cisco executive — Cloud Computing News
Full disclosure: I work at Juniper Networks.
Do you think people will be talking about SDN in 3 or 4 years? Or is it just networking by that time?
IMO, SDN happens to be a name for the moment in time we find ourselves in right now. There are some technologies (OpenFlow being the one getting the most press but by no means the only one). They have been lumped under a neat moniker. And people are writing about them as if they are the thing.
The evolution of workflow automation (or even architectural separation) did not begin with SDN, nor will it end with the current crop of protocols. Until people start talking more about the endgame instead of the tool, there will continue to be confusion.
It’s like talking about hammers. You can do anything with a hammer, so depending on what you want to do, the conversation is different. Start talking about building a house and suddenly the hammer has meaning.
I do think pundits and bloggers, technologists and architects do understand the difference between tools and endgames, but I don’t think the general dialogue reflects this yet.