An edited version of this story ran on Network World on September 7, 2017. Credit: Network World/Computerworld.
The Internet of Things, a vast network of connected microdevices, sensors, small computers, and more throwing off data every second and sometimes even more often, is all around us. Whereas before we had dumb roads, dumb cities, dumb devices, and the like, now all of those things can talk to us and tell us how they are doing, what their current status is, what their environment is like, and even what sort of other devices it knows are near it—and it chatters all of this back to you. All of this is really coming to a head now because sensors are cheap, their processors are cheap, wireless is everywhere and becoming less expensive, and there are tremendous resources for storage and analytics out there. So how do you deal with this phenomenon and take it by the horns to make it begin working for you?
And deal with that you must, because that is coming to you—data. Cisco research projects that there will be 50 billion connected devices by 2020, and all 50 billion of those will be sending off sensor data between once a second and once a minute. How much data is in each of those payloads? Assume it’s 500 bytes, to allow for some packet overhead: that means 25 terabytes of data being spun off somewhere between every second and every minute. What if Cisco is half wrong? What if Cisco is twice right? One thing is for certain: there will be a tremendous amount of data, and with that data comes projections from Gartner, IDC, and Forrester that show a multi-trillion opportunity in both cost savings and new revenue from the IoT.
One other factor is that IT is starting to fade into the background a bit, at least as a central place where technology projects begin. Often business units and individual departments are focusing on technology efforts, made possible by the fact that now there are cloud products out there that let you plunk down a corporate purchasing card and get started in minutes. Hire a data scientist contractor, outfit your kit with some sensors, start up a cloud service account, and before long you could have several terabytes of sensor data and be ready to get fired up. Microsoft has been innovating in this particular area with its Azure service, and they have some compelling offerings.
The Microsoft Azure Internet of Things Story
The Azure IoT Suite is designed to be a quick start type of portal—a true example of a platform as a service (PaaS) offering) that gives you the resources necessary to deal with all of the data being sent back to you while also allowing you to manipulate it, develop some level of understanding of it, and use it to either improve your business processes, solve some nagging problem which you have deployed a boatload of sensors to track and mitigate, or build a model of the way certain parts of your assets behave.
NOTE: there is also a newer software as a service (SaaS) offering called Microsoft IoT Central, which takes the platform level stuff away and focuses only on the software that powers the sensors and connects everything together. Manufacturers can build their own SaaS-based IoT solutions hosted on the Azure IoT cloud service and get their solutions to market more quickly without having to reinvent the plumbing, the platform, and more. There’s also the very new (as in spring 2017) Azure IoT Edge suite, which lets programmers develop logic for small computers and sensors on the “edge” of an IoT environment in convenient languages like Java and C# rather than assembly and other more obscure languages. In this story, however, I will focus on the Azure IoT Suite because I think it more clearly highlights the capabilities of the overall platform.
The Azure IoT Suite itself bundles together a bunch of relevant Microsoft Azure services into a surprisingly simplified package. It starts off by allowing you to create a couple of ready made IoT consumption scenarios, including predictive maintenance and remote monitoring, and automatically orchestrates the various Azure products like databases, websites, web services, the data ingestion point, and more, creating and linking them together so that you are ready to go from square one. For example, for the remote monitoring solution that you can start with as a pre-configured package, Azure self-selects and configures the following services, handling the provisioning process automatically:
- Azure IoT Hub (1 high-frequency unit, also called an S2 unit)
- Azure Stream Analytics (3 streaming units)
- Azure DocumentDB (1 S2 instance)
- Azure Storage (1 GRS standard, 1 LRS standard, 1 RA-GRS standard)
- Azure App Services (2 S1 instances, 2 P1 instances)
- Azure Event Hub (1 basic throughput unit)
Each of the other solutions have a different makeup, but you get the idea: everything you need with just a couple of clicks.
The pitch from Microsoft is that while you might have the internal resources to do a couple of IoT style projects today, as you build on those developments, create new models, deploy more sensors, and in general double down on IoT in the future, you probably will not be able to (at least cost effectively) handle all of that data. You will either be forced to invest in expensive storage infrastructure on premises, or you will have to make problematic choices about what data to keep present, what data to roll up, summarize, and archive, and what data to discard. And of course, when you discard data, you cannot get it back, so you might be losing out on some predictive capability you do not know about yet; if you roll up and summarize data, you lose the granularity and resolution on that data necessarily to build some advanced mathematical models and use machine learning.
Instead, you can start right out in the cloud and take advantage of the tremendous scale of resources that Azure already has—and that is growing quite a bit each year. Instead of spending of disks and compute, you just pay for the Azure services and run times that you consume with your project and you can scale up or scale down as your needs change. Even better, Microsoft is starting to make some of the glue work so you can see the day when your Azure IoT data could be integrated within Power BI, for example, so that your regular knowledge workers (as opposed to trained mathematicians and data scientists) could query your data sets using natural language and get results back in a nice, graphical, easy to consume format. All of that glue and linkage would be much harder to create in a on premises environment, and I think Microsoft here is betting that IoT initiatives are new and green enough in most enterprises that it is not difficult to start them off in the cloud—or at least not as difficult as, say, deciding to move SharePoint into the cloud. In fact, right now, the Azure IoT tools integrate with the Cortana Analytics solution, which provides data science, number crunching, and machine learning tools, and you can then inform your business processes of the insights you derive by linking Cortana Analytics with the Microsoft Dynamics suite of enterprise resource planning (ERP) tools.
Imagine this type of scenario: you operate a fleet of large transportation vehicles, each equipped with two or more really expensive engines. These engines can be instrumented with sensors that report quite a bit of information, including fan speed, oil pressure, ambient temperature, fuel pressure, thrust at any given moment, air quality, vibration, and more. You start collecting this data across the thousands of engines that you have in your fleet and pinpointing that data against maintenance records, failure notices, mechanical delays that interrupt the scheduled service you normally deliver with your fleet, and more. Over time and with some math, you are able to build a model that will be able to show that certain engine components are likely to fail after a certain number of cycles. You can learn which components those are, order those parts in advance, and adjust the fleet deployment schedule so that those parts can be replaced when the equipment is nearby, reducing interruptions and saving the cost of ferrying parts all around your locations. This is the kind of model you can build with Azure IoT Suite (and it happens to be one of the sample models you can run as you get started with your account).
As far as the sensors go, last October Microsoft launched its Azure IoT Suite Device Catalog [https://catalog.azureiotsuite.com/], which showcases more than 500 devices from more than 200 partner manufacturers that are all certified to work with the Azure IoT suite. On the developer and software side, the Azure IoT suite is a full scale member of the Azure service, and thusly works with Visual Studio, Eclipse, Chef, Puppet, GitHub, PowerShell, Python, MongoDB, Hadoop, Ruby, docker, MySql, and anything else that is part of the set of compatible offerings and capabilities with Azure.
How It Works
You can get started by heading over to https://www.azureiotsuite.com and logging in with your Microsoft account. There you can either use your current MSDN Azure benefit or fix up a new one, and then you’ll be presented with the Provisioned Solutions page, which is the first page of the Azure IoT Suite package itself. Then, follow these steps.
- Click Create a new solution to build your own IoT “workspace.”
- You can then choose a couple of different preconfigured solution types, including “connected factory,” “predictive maintenance,” and “remote monitoring.” For this walkthrough, I’ll show you remote monitoring, so click the latter option.
- The Create “remote monitoring” solution screen appears. Here is where you enter a friendly name for this workspace, the Azure region in which all of this should be spun up (you would ideally want the region closest to either you or your sensors to reduce latency), and the Azure subscription to which all of this should be billed. You can find pricing information for all of the components of Azure that the IoT suite will provision at https://azure.microsoft.com/en-us/pricing.
- Click Create solution, and then grab a cup of coffee while Azure spins up all of the resources it outlined.
- After the provisioning is complete, you’ll be back at the Provisioned Solutions screen, and your friendly named workspace will be shown there with a green check mark. Click the Launch button to get inside.
- You’ll be greeted with the dashboard screen. This shows a map of the Seattle area with four sensors geoplotted, each with a colored indicator (green or red). These sensors are simulated, just to give you an idea of the type of dashboard you can build with your own sensor deployment. On the right side, you can see the Device to View menu, which gives you a drop down selector where you can pick individual sensors to examine. On the lower left side, there’s the Alarm History section which shows sensors that are meeting some predefined problem threshold, and then on the lower right you see speedometer looking graphs that show various properties and values that the sensor fleet is reporting.
- On the left side, click Devices. This gives you a grid-style list of devices. You can ue the “+” button in the lower left to add a new sensor, which can be either another simulated device or a physical device with SIM card (ICC ID) for cellular connection, or access to a wireless connection. You can also modify the properties the simulated sensor displays to the workspace, including model and revision number, asset tag, or anything else you like.
- On the left side, click Rules. You can add new rules that operate on the two existing data fields, temperature and humidity, and set the thresholds that apply to those rules. This area is what kicks off those alarm history items on the dashboard, and note that if a device is alarming, its status on the map is changed from green to red to make it easy to identify.
That’s a quick walk around the preconfigured solution, but the key thing to remember is that all of this is live with Azure. You can go and adjust any of this, from the configuration of the dashboard to the way resources talk to each other to anything else; you manage all of this from within the Azure portal, same as any other Azure resource. If you’re looking for a remote monitoring solution just to get started, this solution saves you a lot of effort to get the right pieces in place—start there, tailor it, and build on from there. There’s no additional charge to start here other than the resources the solution spins up to run itself. The design and code is free.
The Last Word
Microsoft has a robust set of tools for integrating all sorts of devices into an IoT solution. They have more scale than you do and work with a wide variety of devices. If you are building an IoT solution, then you owe it to yourself to play around with what Azure IoT can do. I have not seen a solution in this space where it is easier to get started and make your own tweaks to build your own workspace. Worth a look.