Bernie Anger is vice president and general manager of GE’s Intelligent Platforms. We spoke with him in August about the company’s new plans for its industrial equipment: APIs. The plan is to build RESTful APIs for GE industrial equipment, and to give developers a cloud-based IDE in which to build controls for automating and monitoring said equipment.
SD Times: What is the existing business inside of GE that you’re hoping to bring to the cloud?
Bernie Anger: When you think about some of the underlying automation we use in industrial platforms, they exist in two forms. When you think about building a control system, there are two types of controls: one [for] dedicated controls and [another for] general-purpose controls.
An aircraft engine or a turbine has a control algorithm, but we also have a set of products that are general market-automation-type platforms. We call them Intelligent Platforms. We create platforms that are enablers for people who are trying to automate things. An OEM that builds a piece of machinery will use our equipment to automate that machinery.
Generally, we have fairly horizontal products. People will use our control platform to automate different machines.
Dedicated controls are for someone who says, “I want to build something to automate a single process,” like the anti-lock braking system in a car. It does one thing and does it well. These are usually developed using traditional development tools. It might use real-time Linux, and you’d use an IDE to build it.
It’s safe to assume as a corollary that the people doing that work have a good understanding of the constraints of real-time systems. Normally, specialized single-purpose controls are designed by people who understand that. When you cross over into the world of general purpose, we end up isolating customers from the majority of the complexity associated with building real-time systems.
Historically, control systems were designed to last 15 years. The average Intel chip lasts seven years.
Why is it now time to tie these control devices into the cloud?
There is now a lot more computing power at the edge nodes; say, a computer that sits next to the oil wellhead or packaging machine. Take that, give it a level of computing power similar to what you see in a desktop device. Rather than having it be a dedicated piece of equipment, it allows you to combine that with an almost unlimited server in the cloud.
So how do you get these systems to a point where they take advantage of the cloud?
The first stage of cloud for us is to think about a cloud-hosted and cloud-based IDE and operations front end. The IDE facilitates sharing and the use of collaboration.
We do have a more constrained problem set. I can build a set of IDEs for designing and laying out control problems, and it creates simpler IDEs for other problems, and customize the ones you’d want to replace Eclipse or Visual Studio.
Historically in our industry, we’ve always developed IDEs independently. What we’ve done differently this time is build it conceptually like the way Eclipse works. There are layers of plug-ins, like client-side visualization or server-side components. By building those using some standard type formats, languages and REST, it allows us to start leveraging componentry we don’t have to develop.
In addition to being an IDE, it’s also a portal into the operational layer. From the cloud, I can build the controls I want running on those devices. We have a mechanism to go into any device and maintain it remotely, on any machine that can run an HTML5-based browser.
Does this mean making industrial control equipment accessible via the Internet using REST? Isn’t that dangerous?
Security is the No. 1 constraint we battle with on this. When you think about the process of flattening the network, historically control equipment has been protected by not being connected at all. That becomes a process of aggregate transform and protect. Part of the secret sauce the company is working on has to do with how do you create a flat network that’s secure for industrial equipment. You want to be able to use REST-type interfaces on that without exposing yourself to security risks.
Control systems were designed in the pre-Internet era. Most of them become very weak when you connect them to the Internet. The industry has done things to address that. Look at Black Hat Security Conference and you can see that these are definite risk points, and we are very conscious of that.
How close to production is all of this stuff?
The way I would define it at this stage, the announcement is about a platform that allows customers to create, share and control content from a cloud-hosted platform, plus a single point of entry into that content from the cloud, that combines with edge node devices and controllers to be an Internet of machines. You have an edge node device that is Internet-aware, combined with a cloud-based IDE.
We’re actually doing limited demos at what we call innovation showcases around the country. Those started in May. This will be a commercial product in 2013. The way everything goes, when you look at it from afar, it’s exciting, but from a day-to-day perspective, you want it to go faster.
Are customers receptive to this idea?
The level as we take it out to customers, the sequence is almost identical: It all starts with the customer saying, ‘It won’t apply to me,’ then the customers see what it does, then they sign up. It’s fascinating. That process takes three months or 15 minutes. People overcome the fact that this is different, and they see the potential across multiple dimensions.