Originally appeared on SiliconANGLE
Designing an analytics platform for the Internet of Things (IOT) present some big challenges that users of typical business analytics systems don’t have to confront.
Devices in remote locations with slow or nonexistent connections – such as sensors on wind turbines in the North Atlantic – aren’t like office PCs. That means decision-making needs to be pushed closer to the point at which data is collected, and analytics must be distributed between devices, controllers and the cloud.
“The industrial IoT is much more complicated than consumer IoT; you don’t need many intermediate hops for a thermostat in your home, and you don’t need to worry much about the speed of light,” said Marc-Thomas Schmidt (@marctschmidt, above), chief architect of General Electric Co.’s Predix cloud platform and a former distinguished engineer and chief architect of IBM’s Watson platform. Schmidt was interviewed by Jeff Frick (@JeffFrick) of theCUBE, owned by the same company asSiliconANGLE, at GE Digital’s Innovation Day 2016. (* Disclosure below.)
In contrast, industrial environments present a mesh of controllers and aggregation points that have to be coordinated to make decisions in near real time. Add to that the challenges of location, vibration, temperature and power, and the IoT picture gets pretty messy.
Predix is designed to use the entire spectrum of available nodes to aid in decision-making. “Our challenge is to weave all of those compute nodes into a homogeneous kind of system so our customers don’t have to worry about whether they’re working in the cloud or at the edge,” Schmidt said.
Predix is a platform-as-a-service (platform) based upon Pivotal Software Inc.’s Cloud Foundry that is tuned to the unique requirements of industrial applications and the specialized needs of industrial customers, which can span the spectrum from improving operational efficiency to preventing explosions. Each environment is unique, which is why Predix is optimized for adaptability, Schmidt said.
“The edge manager component of Predix knows about all the devices that are out there and is capable of establishing secure connections with each one,” he said. “It then manages the pushing of data from the edge to the cloud and from the cloud to the data center.”
The proof is in the pudding, however, so GE has been on an aggressive push to enlist partners to build applications that illustrate the power of Predix. It’s seeking to apply the platform to “a very popular set of use cases to show what you can do,” Schmidt said.
One of the challenges GE faces is that the industrial market has been under-served by analytics. There is no platform comparable to Predix, and there are few applications yet to show.
So GE is taking a building-block approach. It will introduce some initial applications that solve common problems – such as monitoring equipment for signs of failure – and let customers extend from there.
Other customers want to build their own applications from scratch. “They’ll want a platform to easily build a dashboard on top of all their wind turbines across the world,” Schmidt said. GE will learn from their examples and hard-code the most generally useful applications into micro services that can be used elsewhere.
Ultimately, the power of platforms like Predix will come when customers go beyond predictive analytics to create “proscriptive analytics,” in which historical data can be mined to determine the likelihood of future outcomes. “How do you predict the behavior of machines that hardly ever fail?” Schmidt asked. “It’s hard to find the needle in that big data haystack.” But with data mining on a massive scale in the cloud, patterns will emerge that aren’t apparent today.
The architect is also excited about the prospect of applying data across functions and industries. Combinations will yield breakthrough applications that we can’t even imagine today. “Can sensors in airplanes help make airports and the air traffic system more efficient?” he asked. “As industries becomes more digitized, many things are not that interesting by themselves will become very interested when put together with other things.”
GE wants to serve as a hub for its customers’ IoT analytics activities, sharing best practices and coding the most useful ones into services. It is particularly revved up about the idea of the "digital twin,” which is a virtual version of a physical device.
Digital twins enable customers to experiment with combinations of different information to yield heretofore undiscovered insights. “Don’t just think about putting a dashboard on top of devices; think about creating a twin for each asset you have,” Schmidt said. “Build systems of those twins. You’ll find that devices at the edge can use information from other devices that you didn’t know was useful.”
Watch the full interview below (22:36)