How We’ll Know When the Open Infrastructure Revolution Has Arrived
by Peter Burrows
It’s been four long years since VMware bought Nicira for over a billion dollars. That shocking price-tag—for a company with essentially zero revenues—sounded the starting gun for what was expected to be a rush by vendors to create new, open ways for companies to build, operate and monetize their networks. Rather than be locked into whatever the established hardware vendors happened to be selling, chief information officers would soon be able to cobble together the network of their dreams using that miraculous stuff called software.
Yet most CIOs are still dreaming. Sure, Internet giants such as Facebook and Google have the staff, cash, and clear incentive to roll their own software-defined networks, and a few dozen leading corporations—many of them ONUG members—have followed suit to one degree or another. But while we have seen a proliferation of new vendors, it has not coalesced into a healthy, comprehensive ecosystem of healthy new companies that all those other companies can depend on.
In fact, by far the biggest new trend in networking is the acceptance of the public cloud as a way to effectively outsource networking, mostly to giants such as Amazon, Microsoft, and Google. According to a CIO survey by JPMorgan from earlier this year, 40 percent of workloads will move to the public cloud over the next five years. That’s provided a great new option for CIOs that want to cut costs and speed time to market, but gets most companies no closer to the dream of being able to use networking as their own competitive advantage. Just as customers have carped over being locked into OEMs such as Cisco over the years, now companies are faced with being locked into AWS or Azure.
The slow progress is evident in the slowly changing spending patterns. According to Infonetics, Cisco has lost three points of market share in the three years, but the top five makers of switches – Cisco, HPE, Huawei, Juniper, and Arista—have retained 83 percent of the spending on Ethernet switches. So-called white boxes had market share of only 2 percent in 2015. While HPE, Dell, and Juniper have created their own hardware capable of running other companies software—so-called Brite Boxes—these machines have made only a small dent in the marketplace, says Infonetics’ Cliff Grossner.
I’m not discounting all the good work that’s been done by ONUG, and by individual companies that have insisted on a more open, software-centric approach (this includes big service providers such as AT&T, which intends to virtualize 75 percent of its network by 2020, up from 30 percent this year). The pressure has clearly had an impact. Cisco has clearly gotten the message, the evidence being its sustained push behind its ACI architecture. True, ACI is an attempt to deliver many of the benefits of open networking in a way that still leads to sale of its proprietary hardware. Yet to accomplish this, it has had to open up far more than in the past – for example, by exposing APIs so other infrastructure software companies can create applications to run on Cisco-based networks.
These APIs are one sign that conditions are improving for development of an honest-to-goodness open networking ecosystem. There are still plenty of innovative start-ups scrambling to position themselves as leaders should that time arrive. Apstra, for example, is gaining traction with a distributed operating system that takes advantages of those APIs to let operators more effectively monitor and manage what’s actually happening in their networks. SnapRoute, a younger start-up created by former IT infrastructure managers at Apple, is developing an open-source networking stack that could potentially do for switches what Linux did for servers. Other promising companies such as Thousand Eyes and Forward Networks are creating big data analytics applications, to make better use of the detailed traffic information. Even Big Switch, an early SDN pioneer that has struggled to find a solid business model, is making hay with this approach. And a generation of cloud-companies, including SD-WAN providers like, Viptela, VeloCloud, and CloudGenix, are creating ways for businesses—even smaller businesses without IT staff—to rent infrastructure suited for their needs rather than be limited to commodity offerings from Amazon and the other cloud giants. Optical networking is also looking strong, evidenced by the success of companies such as Acacia Communications.
And ONUG is pressing ahead to address other weak points in the open networking ecosystem. It is defining use cases that its members can use to pressure vendors to develop the necessary products and features. With their combined massive buying power, this will redirect some of the $1B spent each year on infrastructure tools towards open alternatives. Clearly, most companies are going to have a mix of open and closed software, and will use various cloud services along with homegrown applications. But while vendors’ proprietary software won’t disappear, an open, vendor-independent control plane offers the best approach to get all of these disparate products to work together.
That’s all necessary, but the real sign of success will be the arrival of the next generation of successful companies—not just nutty acquihires at bubbly valuations a la Nicira, but companies with sustainable, highly-profitable business models. It’s not a precise proxy for open-networking, but Nutanix’ blockbuster IPO – the stock jumped 178 percent in its first two days of trading – indicates that there are plenty of businesses out looking for a new way to build their physical networks
Maybe the best indicator that the revolution may yet occur is the continued optimism of entrepreneurs. “Our success couldn’t have happened five years ago,” says Apstra CEO Mansour Karam. “But five years from now, it will be too late.”
Peter Burrows is a freelance technology journalist and content creator. During a 22 year run at Businessweek and Bloomberg News, he covered marquee companies including Apple, Cisco, Hewlett-Packard andMicrosoft. Many of his hundreds of stories dealt with the evolutio of enterprise computing and carrier networks, including the rise of the Internet in the 1990s, cloud computing in the 2000s and SDN and NFV this decade. Burrows wrote “Backfire: Carly Fiorina’s High Stakes Battle for the Soul of Hewlett-Packard”, the unauthorized story of Fiorina’s failed tenure at HP, and co-wrote Businessweek’s award-winning memorial issue to Steve Jobs in 2011. Burrows has appeared on PBS News Hour, Charlie Rose and Bloomberg West. Since leaving Bloomberg in 2016, he has written for publications including MIT’s Technology Review, and provided consulting and writing services to a variety of tech-related firms.