IoT Main

IoT as a Service with Azure

By Rodney Guzman, CEO/CDO/Co-Founder/Owner at InterKnowlogy

Do you have IoT envy? Many SMB’s are not rolling out IoT because of complexity and cost. Talent to do IoT “right” without flailing about is difficult to find. A successful implementation requires skills across many disciplines. On the flipside, companies with many IoT dog years under their belts are struggling with the decisions they made a long time ago. Their growth is outpacing their ability to scale, data storage problems are rampant, reporting is hardcoded and difficult to change, and there is no ability to glean any insights from their Big Data.

Fortunately, cloud platforms are maturing rapidly in this space.

If you already have expertise with these cloud platforms, then building a complete IoT system can be done in a couple of months. To help guide our customers through this process, there are a short list of considerations we typically focus on to get started. These include:

Ingest. You can have thousands or millions of IoT devices. They can be deployed regionally or worldwide. The IoT devices might be chatty and produce many messages. There are likely dependencies where exponentially more messages will be produced during certain conditions, such as the time of day. The system you use to ingest all of this data should be just as flexible as your IoT devices. You do not want to be paying for services you are not using (such as when the IoT devices do not send frequent messages). Your data ingest should be near your sensors, and you might need to deploy your data ingest to multiple regions or even globally.

Real-time Analytics. Your IoT devices can be producing millions of messages every minute. What is happening at this very second should be extremely important to you. The temporal nature of the data makes it challenging to report on. Your cloud platform should have the capabilities to do this heavy lifting for you so that all you have to decide on is what you want to learn from your flowing data. You should not have to worry about managing your data to prepare it for this type of reporting – let your cloud platform do that magic for you.

Storage. There is a science to selecting the right type of storage. It is important to understand the shape, record size, and volume of data flow to determine what storage technology to use. This was probably the first thing that every cloud platform nailed – provide enough flexibility in data storage technologies that you can pick the right one that meets your needs.

Reporting. Integrated reporting is extremely important. It needs to easily connect to any data storage technology you are using. Adding and customizing reports should not require a data scientist. Standing up a dashboard should be measured in minutes and hours and not days and weeks. The reporting technology should be ubiquitous and accessible anywhere (phone + desktop), and be able to easily report on all the different storage types. The real-time data ingest process should feed right into the same reporting engine.

Bidirectional. Not all IoT devices are just data producers. Your thermostat, for example, can receive remote control messages. However, there are instances where your IoT devices are not singular and need to work in harmony with each other. This type of orchestration is complex. The logic and control needs to be externalized from each IoT device and centrally managed. This plumbing will serve multiple roles, including discretely managing each device while virtualizing the logic that orchestrates multiple devices. Being able to codify this intelligence easily is critical to control complexity.

Insights. Having the tools to build models for machine learning is no longer considered a nice to have. Even if you do not know what to do with machine learning, knowing that it eventually can be done is important. Collecting the right kind of data now so that machine learning can be done when you need it is critical. The tooling should enable you to find correlations in your data, meld together data from various sources, employ different types of analysis, and allow you to form predictions. These tools allow you to codify your domain expertise and to automate decision-making. For example, after training your models you have essentially codified what might have been expert knowledge that only exists in peoples’ heads. The cloud platform should easily wrap that predicative knowledge in a service that you can integrate throughout your apps.

Data Factories. New data is coming in all the time. Once you train your machine learning models, you want to make them better and better over time. You need to consider how new data will affect the models. The facilities to take new data into your models needs to be part of the cloud feature set.

For all the above, we use Azure.

The tools Azure provides enable us to make quick work of an IoT challenge. In fact, the tooling has become so good that it has enabled new opportunities for us to consider as a business and expanded how IoT can make a difference for our customers. By way of example, the following is one area we believe we can make difference because of Azure:

We are envisioning an IoT as a Service platform that will consume IoT data from thousands of different device types.

By working closely with new IoT chip manufacturers, this platform will be a “first class citizen” in their SDK’s. The devices we will be targeting are those that produce data because of a consumer interaction. For example, if someone is doing laundry and interacts with their washing machine, the washing machine can report that interaction via IoT. In just three short years, there are some who are projecting a full 8x increase in IoT deployment. More and more devices a consumer interacts will be online.

Today, when a consumer is profiled, it is typically done from their interactions with software. This includes web sites, mobile apps, etc. What if that consumer profile can be broadened with IoT data? Everything that a consumer interacts with, whether software or hardware, could be part of their profile. What could be learned about consumer behavior that we could not have known before? Whether we like it or not, and most likely not, this is the level of tracking that we will all be eventually subjected to.

If you like to learn more about how Azure can empower your IoT initiatives, please contact us at

Tagged on: