A month ago Benjamin Black, one of the creators of Amazon’s Elastic Compute Cloud, joined Pivotal, adding to the ranks of big-name cloud computing stars that have recently joined the EMC (EMC) spin-off. Much was made of Black’s bona fides, but no one talked about his mission. This was odd, since he’s tackling one of the most challenging and important pieces of technology infrastructure around today: building an orchestration layer for the so-called Internet of things.
Black is part of Pivotal’s Cloud Foundry team, which builds an open source software-as-a-service framework for big-name companies and tries to teach companies how to use that software to build their own infrastructure. If Black succeeds in his efforts to build a scalable architecture for the Internet of things, Pivotal will go up against companies both familiar to it and new. It will rival IBM’s (IBM) IoT Foundations project; efforts at Citrix (CTXS), which in December bought Octoblu; and manufacturing software provider PTC (PTC), which has purchased ThingWorx and Axeda to build an Internet of things cloud platform for industrial use.
And yes, there are dozens of other big-name companies from Cisco (CSCO) to Salesforce (CRM) that are also trying to market their way into this sector with varying levels of success and actual products. There are also startups, many of whom are the targets of those same big name providers seeking an entry point into the business. So why is Black special? Aside from his role co-authoring the paper that led to the creation of Amazon’s cloud services, he has managed to articulate in a way that few others have, the essential challenge that makes building a scaled out infrastructure for the Internet of things so different from his days building out a scalable infrastructure to manage servers.
Unlike virtualized servers, which when configured into cloud deployments can be turned off and on without affecting the app running on top of the computer, sensors aren’t fungible. When you lose a sensor, you lose integral information—another sensor can’t just take over its job. This is the same for actuators, which are the connected devices that take action, such as locks, levers, or switches. Thus when you build a system to manage a system of sensors and actuators, you have very different constraints than when you build a system for managing a system of servers.
That’s what also makes this an area that’s wide open for startups and big companies. Anyone could invent a core technology or way of solving the problems associated with managing physical resources that are stubbornly anchored in the real world and can’t be virtualized. This may seem obvious, but when it comes to deploying software to manage scaled-out IoT systems, very few companies are dealing with these problems in a repeatable way. That’s why Black is so excited.
Beginning late last year a large number of customers started coming to Pivotal having trouble scaling their real-time event processing efforts. Black said that these customers might be dealing with industrial processes or crop monitoring, but they all had issues they couldn’t solve with the software they were finding on the market. Most of it was for older systems or aimed at helping developers get prototype devices online.
Many of these new efforts, such as Octoblu, which was designed by former Bechtel engineers, aimed to start with the developer side and scale up for the industrial customer, but Black wants to start big and stay there. Given that Pivotal has backing from GE (GE), which is also a big proponent of the industrial Internet of things, the push to go big makes sense. But there may be a reason that other firms—even IBM—have started small. Scaling out an orchestration platform that can handle thousands or tends of thousands of edge nodes in the form of sensors and actuators isn’t easy.
For now Black is envisioning an architecture that uses a gateway on the factory floor (or in the field) to act as the hub for the sensors and to aggregate information that will then be sent back up to the cloud. Not every instance has to have this architecture, but it’s the easiest way to think of it for now. The essential ingredient is the orchestration layer that will let a developer manage how changes in the software are applied across the sensor layer.
A developer should be able to register new devices on the system easily, roll back code changes if they cause problems, update firmware on classes of devices or select devices, address security and monitor and manage the status of each sensor. Black isn’t sure if he wants to use the existing protocols people are trying to implement for the industrial Internet such as MQTT or XMPP or if he will try to repurpose web protocols such as HTTP/2.
Whatever he decides, he has to do it quickly, because he says Pivotal wants to announce some progress on his orchestration layer this year.
Watch more business news from Fortune: