In April 2013, Plexxi teamed up with SDNCentral to take a look at how the SDN market might emerge. The original post along with supporting info graphic and written analysis can be found here. At a high level, the major takeaway was that we predicted that between 30 and 40 percent of the networking market would be influenced by SDN by 2018. At the time, this was BY FAR the most aggressive take on SDN. IDC had been projecting a little more than $3B by 2018, which would put their estimates somewhere around 5% of the overall networking spend.
So 18 months later, how do I feel about the analysis?
SDN spend is largely substitutive
In the original analysis, I made the point that SDN spend is not likely to be net-new dollars coming into the networking industry but rather a shift in dollars from traditional networking equipment to SDN-enabled equipment.
How’d I do? I’d say that this was spot on. Of course, this was the easiest of the predictions at the time. It is rare that dollars are created; they are usually shifted from somewhere else. Here, all I was really predicting was that the somewhere else was other networking spend.
I did make the case for some net-new spend on the controller side. In the analysis, we predicted that the controller and orchestration market would top $1B by 2018. When I wrote this, I was thinking that there would be some spend on companies like Big Switch but that the bulk of the revenue in this space would go to VMWare (by way of the Nicira acquisition).
If you believe the Nicira bulls, it looks like their revenue could top $500M this year, which would put them on pace to take out the $1.2B target before we add any of the others (PLUMgrid, Midokura, and so on). That said, Nicira is selling more as a distributed firewall, which seems a little bit unfair. If they eclipse the $1.2B by themselves, I am going to claim I was robbed.
SDN will hit all areas of the network, but datacenter will be tops
In the original analysis, we stated that virtually every aspect of networking would be impacted by SDN in some way, but the datacenter stood to see the most action.
How’d I do? This one seems mostly correct. At the time, most of the SDN startups were focused on the switching space, and the incumbent activity was targeting a mix of datacenter and WAN. If you look at ACI and NSX, that action is all in the datacenter. But efforts like Juniper’s OpenContrail and startups like vIPtela show that there is activity outside the datacenter.
Revenue is still in the early days, so it’s tough to declare an outright win here, but the signs are promising.
Skill sets were going to act as a drag on adoption
We predicted the rise of more programming-type networking jobs. To be fair, this was already underway, so it’s hard to take too much credit for being right on this one. We further claimed that the lack of skills would slow adoption.
How’d I do? This is almost like predicting the sun will come up. Yes, the analysis was correct.
DevOps has certainly gotten more attention in the last 18 months, helped in part by the promise of network programmability and in part by the rise of orchestration frameworks like OpenStack and all the tools that surround it. The other part of the analysis—that this would act as a drag on actual deployments—appears to have also been mostly correct. There are SDN training and certifications just now becoming available.
Professional services will become increasingly important
As part of the analysis, we pointed out that integrating all these technologies was going to put a premium on professional services. We predicted that professional services practices would rise up to fill the gap, providing competitive advantage for companies who do this well.
How’d I do? On the surface, I still want to say I can be right, but objectively, there just isn’t a booming professional services business around SDN yet. Some resellers have jumped in, but they aren’t really acting like professional services companies so much as just expanding their portfolio to seize on the market buzz. Oracle hasn’t dipped in as much as I thought they would, and Cisco isn’t touting services despite an internal effort to develop the practice. The reality is that services is still not emerging.
Now, it could be that it is still early days. But if I am being honest, I have to declare defeat on this one.
SDN purchase decisions will lead actual deployments by as many as 2 years
The headline of the analysis was SDN market to reach $35B by 2018. But the real report actually said that this impact was more about purchase criteria than actual deployment. We had stated that deployments would lag purchases by several years, but we expected SDN to become a standard part of the RFP and bidding process for most large deployments.
How’d I do? Well, deployments of SDN are still very few and far between. So on that part, the analysis appears to be correct. And it is also true that SDN support is beginning to make its way into more and more RFPs. It’s hard to know whether the deployments are going to happen en masse in the next year or so, but this prediction appears to be holding up.
The Big One: SDN will impact $25-35B in networking spend by 2018
The biggest prediction was the market size. So can we stand by the assertion that 30-40% of networking will be impacted by SDN by 2018?
The jury is still out. Every vendor in the space is now pushing SDN. Cisco’s new switches have SDN support, and they are porting it back to the other Nexus devices. This would leave the already-long-in-the-tooth Catalyst line as the lone standout on the switching side. Juniper is the one bucking the trend here for me. They are using Contrail, which means their overlay could keep them from having to actively sell SDN across their portfolio. So while their solutions are SDN-enabled, their individual devices might not be. Every HP switch supports OpenFlow. Brocade is very aggressively pursuing both SDN and NFV.
I challenge you to find a set of meaningful devices NOT supporting SDN at this point. It would appear that the market is going to tip. Now, the real question is whether people buy because of SDN or not. Buying something that supports SDN is not the same as having SDN influence your decision. Every car has cup holders, but I don’t use cup holders as a decision-making criterion.
Here, I think it’s tough to tell. As we are in the trough of disillusionment on SDN, will we see demand rise again? Probably. Will it mean that 30-40% of all decisions factor SDN in? That seems a little bit tough. We are still looking for the killer application. We are still waiting for reference architectures that can be stamped out for the masses. Without those things, it’s hard to see 30% of all networking decisions including SDN, especially when the SMB space cannot even spell SDN yet.
But there is still 3 years. And if every product supports SDN, then it could be that it becomes a de facto driver of business. So I will say it’s too early to call at this point.
The bottom line
The real point of the market sizing exercise was to predict that SDN is going to be huge. Since that time, every major vendor has built SDN into their roadmap (or at least into their marketing collateral). Yes, deployments are lagging as people figure out just what to do with the technology, but there is no doubt that the competitive landscape has been forever changed. We are still in the relatively early days of SDN, so there is a long way to go, but the future is bright.