7 Things to Consider When Making IoT Infrastructure Decisions

By Yevgeniy Sverdlik

The abstract three-word buzz-phrase that describes the concept can mask the messiness and complexity of what the Internet of Things is. It’s a catch-all description for countless different use cases in every possible industry vertical. Needless to say, nobody quite has IoT figured out.

However, at this point, after several years of development in the space, some basic questions and trends have emerged which at a minimum apply to big sections of the market. Last week at IoT World in Santa Clara, California, experts highlighted those questions and shared their opinions on how they might be answered.

Here is just a few of those issues, distilled from multiple expert panels and individual presentations at the show, starting with something very basic but nonetheless consequential:

1. Reuse infrastructure

If you have the luxury of planning out an IoT deployment in a building before it’s built, or at least before the contractor puts in communications infrastructure, such as fiber cables and network switches, consider having the IoT vendor work together with the contractor to use the same infrastructure for the IoT deployment.

Too often do buildings end up with “double fiber, double switches, double everything,” Erik Ubels, CTO of OVG Real Estate, which designed the Deloitte office building in Amsterdam considered the world’s smartest and most sustainable building, said. “It’s the same technology. It’s all IP in the end.”

2. Edge computing isn’t always necessary

The same Deloitte building manages to be the world’s “smartest” without any edge computing, according to Ubels. Data from all the coffee machines, sensors, doors, elevators, power distribution, and so on – in all, 28,000 sensors across 40,000 square meters – is processed entirely in Microsoft Azure, via the Azure IoT solution.

There’s not a single server or VM running in the building to process IoT data, he said. “It all runs in the cloud.”

3. Fluid compute

While there will be cases that don’t require compute muscle at the edge, fluidity in where in the IoT network data processing happens appears to be where things are headed.

More and more computing capabilities are being designed into smaller and smaller devices for IoT deployments. You can already have an edge router that can analyze video feeds from multiple CCTV cameras, for example. Microsoft already has a version of its Azure IoT Edge software stack that can run on a Raspberry Pi 3, including services like IoT Hub, Machine Learning, Stream Analytics, Functions, and Cognitive Services.

And work is underway to push more intelligence down to the end devices themselves. “You’ll see basically more and more of this intelligence pushed to the edge,” Jesse DeMesa, strategy partner at the industrial IoT-focused venture capital firm Momenta Partners, said.

Microservices, enabled by containers and container orchestration platforms like Kubernetes, make it possible for the same IoT software to run across multiple types of compute infrastructure, essentially making possible the intelligent edge itself.

What all this ultimately means is that the future IoT network may be intelligent enough to select the most logical place in the chain a certain type of data should be processed. Some will be processed on the spot, by the IoT device, some at the local edge (by a router, for example, or an edge-computing cluster), some at the telco edge in a colocation data center nearby, and some will be sent to the cloud.

“We basically want to be fluid in where we want that compute to take place.” DeMesa said.

4. IoT data management lagging behind

Microservices are making the intelligent edge possible and so is the trend toward more compute power in more IoT devices. What’s still behind, however, is data management, the mechanism that decides what data should be processed at the edge, what data should be sent to the cloud, what data should be discarded, and so on.

“The application logic has gotten there,” Stephen Goldberg, CEO at HarperDB, said. “The data-management side is really where we need to see more evolution.”

Today, much of the data-management functionality is custom-built for each individual IoT project. “That becomes very expensive,” he said. In other words, some standardization in this area will go a long way in bringing down the cost of IoT.

5. Who owns the data?

This is one of the biggest questions your organization will have to answer before a major IoT deployment. Many of the vendors you’ll be working with will expect to derive a lot of value from the data their devices and platforms collect when deployed on customer premises.

For example, a manufacturing-equipment maker will typically want to keep operational data coming off the sensors on that equipment after it’s installed in a plant. The plant operator will also want the data, so it can analyze it and optimize its operations to reduce cost and/or boost production. And they may or may not want to share that data with the vendor.

“Of course we think that any data we collect off a machine is our data,” Dave Rauch, senior VP for worldwide manufacturing operations at the storage-hardware producer Western Digital, said. There’s potentially an advantage in sharing the data with the machine manufacturer, “but there should be nothing that precludes us from using that data to optimize our factories.”

JoAnna Sohovich, CEO of Chamberlain Group, maker of residential and commercial door operators and gate entry systems, said IoT data ownership is “a giant pricing discussion,” implying essentially that the customer may have to consider a tradeoff between having the benefit of advanced analytics capabilities at relatively low cost and exclusive control of the data.

An industrial customer should decide how much they value having exclusive rights to their operational data versus the value of their equipment vendor using that data to improve their products and services. It’s important to keep in mind that when a vendor invests in data collection and analytics capabilities in their products, potential value of the data they expect to receive is a big part of the decision.

6. Data governance

What’s hopefully clear by now is that much of the IoT infrastructure strategy discussion is about what data lives where. And the question of physical data location goes well beyond the IoT space, especially today, when GDPR and similar new rules in individual nations outside of Europe are going into effect, tightening regulations about where private data can and cannot be stored.

That adds an interesting angle to the IoT infrastructure discussion. It’s easier to comply with such regulations when you’re processing data at the source instead of moving it to a centralized data center, Christos Kolias, principal research scientist at the French telecommunications giant Orange, said.

Under such rules, storing and processing data in the cloud may not be easy when you have international clients, whose data originates in different countries, subject to different laws. Therefore, theoretically, compliance should be easier if you’re crunching a client’s data where the client is.

7. No one-stop shops

If you’re embarking on an ambitious IoT project, prepare to deal with multiple vendors. IoT is an ecosystem play, and you will have to work with multiple layers of the ecosystem, be they cloud providers, system integrators, hardware suppliers, or software makers.

“You never have a one-stop shop,” Philippe Fremont, Europe VP of IoT for the electronics distributor Avnet, said. “You will never have a successful IoT project with one partner. You will always have a few that will help you and guide you.”