Wind River has formally completed the buyout that separates it from Intel, and once again the 37-year-old company is on its own. I spoke with President and CEO Jim Douglas to understand how the company, once a scion of the industrial and embedded software world, plans to adapt itself to a future that’s radically different than what Wind River started out in.
Douglas gets it. His vision for Wind River is audacious. The company began life building software for industrial and embedded customers that would control custom-build machines designed to perform a single function. It’s now pushing its industrial customers to learn from the cloud computing world, so they can adapt to the scale needed for a world with billions of connected sensors.
This is a challenge that Microsoft and Amazon are tackling as well as startups such as Resin.io and Zededa. I’m constantly searching for the new architectures we’ll use to build out the internet of things. I believe that computing will need to be decentralized and happen at the edge, for a number of reasons. I’m also certain that in the near term there will be a lot of vertical specialization; asset tracking will require something different than video processing and industrial shops will have different needs than a retailer, for example.
What’s unclear is what will take place in the cloud, how things will be orchestrated between the edge and the cloud, and how companies will ensure the security and consistency of data as it travels between the two. The traditional IT guys are providing products such as AWS Greengrass or Azure IoT Edge to link the sensors and servers on factory floors to their services in the cloud. Wind River has a similar plan.
First off, Douglas says the industrial guys need to stop thinking of their computing projects as pets and start thinking of them like cattle. In industrial processes, computers are built to perform one job and do it well. They are pets. But if you want to start building applications that can harness the data from a lot of pets and communicate with them, you need to treat your pets like cattle. Cattle is how big cloud vendors treat their servers.
At a practical level, this means Douglas wants to bring the tools of scaled-out compute to the once one-off machines that populate factory floors or even computing systems in planes and cars. Wind River is building tools that will enable virtualization on factory servers and containers that will run on sensors and industrial process controllers. The idea is to create a unified environment that can run applications that are managed from a single point.
If this sounds familiar, that’s because it’s akin to what happened with computing after hypervisors were invented and what is currently happening in the telecommunications world with software-defined networking. This change has both technical and economic ramifications, and Douglas is aware of both.
On the technical side, Wind River is building a product that’s not yet named, but will launch in the fall to help with orchestration between the cloud, the edge, and the sensors. Unlike some of the cloud vendor options, he’s trying to build an orchestration layer that can allocate where jobs run based on critical issues such as latency or reliability. In the data center, a server can go offline with little consequence, but an application bringing data from a machine can’t fail without risking the production process. Wind River’s edge is that it understands those industrial processes better than the IT guys, and can thus build the tools needed to fit their needs.
Then there’s the economic aspects of this shift. When cloud computing arrived it didn’t utterly demolish Dell, HP, or some of the other big providers of servers, but it did change their customer base and their margins. In software-defined networking, the shift is currently causing a world of pain for Cisco and Juniper as their customers start spending more on servers that run specialty software as opposed to Cisco’s specialty boxes that also run proprietary software.
Vendors providing hardware such as industrial process controllers and specialty plant automation gear could soon find themselves trying to move all of their value — which up until now has been encapsulated in the hardware — to software that can run on generic servers. At the same time, companies that have expertise in the actual jobs being done with those industrial controllers will compete to make software that can now run on general hardware.
Wind River hopes its software will enable this future and make the divide between the IT folks and the operational technology folks a little less clear.