Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.
Azure and IoT announcements peppered Microsoft Build 2019, including Azure IoT Edge, Azure SQL Database Edge, an AI and robotics toolkit that uses Azure tools, Azure-related Kubernetes updates, and more. To put the pieces together into a more complete picture of the company’s IoT landscape, VentureBeat spoke with Sam George, Microsoft’s head of Azure IoT.
We started by asking George to give us an overview of where things stand for Microsoft and Azure IoT today.
Sam George: In the early years of IoT, we spent a lot of time focused on making IoT possible — especially for the early adopters, since that’s the phase of the market. We see the market right now moving between the early adopters and the early majority. And so it’s starting to move across the chasm.
A lot of what we focused on over the last couple years has been making IoT easy. You already have a broad set of IoT capabilities — everything from tiny microcontrollers, all the way up to Azure stack, and everywhere in between. We have large set of platform offerings for connecting and managing devices for data and analytics, IoT data business process integration.
MetaBeat will bring together thought leaders to give guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.
Some of the more interesting things that we’ve been doing over the last couple years is providing faster time to value for partners and for customers. We started with solution accelerators. So you could provision a working end to end IoT solution in your Azure subscription in just a few minutes.
We now have a software as a service offering for partners and customers, as well. And that’s based on what we’ve learned over the early years in IoT, which is that a large percentage of these IoT solutions, sort of — you know, they do the same thing, right? They’re monitoring devices to learn something about what’s happening in a business in the physical world, they’re storing data, they’re finding insights over it, and then they’re triggering business processes.
And so that that insight really enabled us to provide a SaaS offering to really take care of the common needs in IoT. Today, we call that SaaS offering IoT Central. In IoT Central, you can push a button [and] provision an app — [it] takes about 15 seconds. It’s a configuration system.
And it takes just an hour or so to customize it for a device or to configure it for IoT Central. And then the nice thing is, as you connect devices to that, we automatically scale the system. We wear the pager, we keep the whole thing running — both the application as well as all the underlying services.
A lot of early IoT, you spent a lot of time building an operational dashboard, and setting up monitoring rules and things like that. And so we made all that easy.
The thing we announced this week is something we call IoT Plug and Play. Because while we’ve made the cloud really fast, the device is still slow. And what I mean by that is even though I can get to a configuration system in the cloud with IoT Central, I’m still going to go write a bunch of code on the device, right? And, effectively, there’s a tight coupling between that code and what the solution expects to receive.
So for example, if the solution is expecting vibration, and humidity, and accelerometer values and things like that, I have to make sure that the device is sending that exact data in that exact format. And so there’s this tight coupling between the two.
You saw this many, many, many years ago, before Windows had plug-and-play; device providers would have to build a lot of code, a lot of software on the device. It also had to provide software in Windows, and the two would have to match. And that’s a lot like what IoT is like today.
The bigger difference is that, number one, it’s open source, it’s cross platform, it works on any operating system, [it’s] multi-language, and all that. So that’s a big step forward. And so I showed a demo in my talk yesterday, where you take a device that Azure IoT Central has never seen [and] turn it on. It gets connected IoT Central. It appears. It starts sending. It starts collecting data. That’s it. And then you just set a few monitoring rules, and you’re doing IoT. So you know, that’s the speed at which we’ve been helping customers get to a production-grade IoT solution.
VentureBeat: How has the rise of the intelligent edge affected what you guys have to do as an IoT group?
George: We saw this turn on the road several years back. And it was the reason why we introduced our edge support with Azure IoT Edge. It’s already generally available, customers are already in production — it’s doing great.
I’ll tell the story from the point of view of a customer that started with IoT and then evolved to edge. That’s kind of interesting, because that really speaks to a little bit of the evolution that we’ve seen over the last couple years. So a couple years back, when we announced IoT Edge, we talked about a customer that was using Azure IoT — Sandvik Coromant, they produce metal-cutting machines. They had developed — in the cloud — some machine learning model that would predict whether or not there was a pending failure on the machine, like the machine was going to get damaged. And when they got that working, the part they worried about was, well, now what happens if there’s a local network outage. [Now] the machine is vulnerable, because we don’t know when to shut it down.
And what’s funny is that they hadn’t known for decades, but now that they did know, they wanted that to run all the time. What they did, initially, is they built a second implementation on the device itself. When we looked at that — you know, we were starting to talk about edge computing — we looked at that as, “Why shouldn’t our customers be able to run Azure services or their own services anywhere they want, whether out in the physical world or in Azure?” And that to us was really part of the genesis for IoT edge.
And so fast forward today, they’re able to take that same machine learning model, run it down on a device, and have it work — we just added in definite offline support. So you can connect, download a deployment. Sever the [internet] connection? It keeps running.
VentureBeat: You can update it, but it’s so purpose-built that you almost don’t need up to update it?
George: But you do, though, that’s the thing. That’s a fascinating thing — sort of the second part of the story. So the pattern used to look like: IoT device sends data to the cloud, finds insights from it, build machine learning model, predict something, right? Now, what it looks like is, it picks up where that machine learning model left off.
Now […] I deploy it out to the device, and it operates at very low latency, tolerance of network outages and all that. But then I periodically upload more data to the cloud, and retrain models, because machines change over time. So something that predicted a machine failure today won’t in two years, because the machine characteristics change. So it’s actually a loop between the intelligent edge and the intelligent cloud.
VentureBeat: A virtuous cycle.
George: That’s right, yeah. It really is. And in fact, the second part of what I demonstrated is an IoT edge device that’s connected to IoT Central. And we made it plug-and-play enabled, as well. So when it connects, it streams video with AI overlays.
I showed a worker safety scenario, where you have a camera in an OSHA environment, where you want to ensure people are wearing hard hats for safety, and it can detect whether or not someone is, and then help remind them it’s time to put on the hard hat.
VentureBeat: How easy is it to [implement IoT solutions]? Because some things do require coding…
George: Cognitive Services are — sort of [the level of the] average knowledge worker can do [it].
We did a demo — I think it was last Build — with Scott Guthrie in our keynote. We had a little IoT edge device that could run one of these cognitive services. And we built a machine learning model using cognitive services that we called “Scott or not?” So there’s a whole bunch of pictures of Scott and then a bunch of pictures of other people. And then we labeled them. There’s a little button and on a website, you click “Build a machine learning model.” And it took a minute or two. And then we downloaded that as a container and deployed it.
VentureBeat: So the basic function of it is easy enough. But what about actually creating useful things? How complicated does it get? How quickly does it get complicated?
George: There’s a large gap between, you know, “I built a cognitive service” and “I built something that can win chess.” Because there’s a spectrum in AI, but what’s what’s been fascinating to see how much you can do things like the cognitive services. And it’s not just custom vision, it’s also speech and language understanding, and text, and, you know, optical character recognition … there’s a lot you can do with it.
VentureBeat: I guess I’ve often thought of intelligent edge and AI as very much one and the same, to an extent. But they’re a little bit different.
George: When I think about these new waves of computing that are helping customers transform, a customer looks at, “How can I improve my business?” And then there’s a set of techniques that enable them to do that — IoT, edge, AI, they’re all very interrelated. So I don’t see them as discrete areas, I see them all in support of helping businesses transform. I really do.
As an example, IoT is a prerequisite, in many cases, for edge computing. You need to be able to manage the device and talk to sensors. Then you can start landing workloads on it. So I can start making sense of that data locally. And then, once I have edge working, then I want to deploy AI to it, and take advantage of the functionality it provides there.
It’s a layered, composed stack.There are these waves of computing that are all happening, roughly the same time, offset only by a little. And it’s providing really significant game changers for businesses.
VentureBeat: In terms of the penetration of the potential for these layered technologies, how far are we?
George: What I can say is that we are really seeing the market moving between those early adopters and early majority. The types of customers that we talked to today are remarkably different than the ones that we were talking to even just two years ago. And they’re coming looking for very turnkey things, or low barrier to entry for them to do it themselves, versus a lot of the earlier customers were sort of the pioneers and early innovators.
I’d say it’s definitely the innovators and the early adopters that are in production right now. And there’s a very large volume of them that have figured out this is going to have a significant benefit [to] my business. And it has.
VentureBeat: Is it mostly enterprise customers right now?
George: The early, early years of IoT, we saw many enterprise customers. But now that we’re making that easier, we’re starting to see more of the small- to medium-sized businesses showing up for sure. And also consumer.