Before you can make enterprise IoT successful, build a strong infrastructure — here’s how

That you’re reading this article confirms that you’re looking for opportunities of empowering and enabling your enterprise to reap the promised benefits of IoT. Right from measuring temperature, pressure, and vibrations in industrial machinery parts for indicating preventive maintenance to enabling unmanned flying vehicles for covering thousands of miles without wavering from the areal course, IoT is a magical world of amazing possibilities that can help businesses of all stripes reach a pedestal they have never been at before and could probably not reach without this amazing technology. However, there is a lot of noise around the infrastructure challenges that IoT will invoke, and this noise must be silenced before you can make enterprise IoT successful.

Make enterprise IoT successful: The infrastructure puzzle

IoT is often oversimplified as “objects connected to Internet.” To the layman, the definition helps conjure up images of what IoT is like. However, to the IT decision maker in an enterprise, it doesn’t capture the fact that the kind of data processing and network infrastructure that IoT applications demands is not anything like conventional Internet-powered applications.

Then, there are several industry-specific IoT applications that pose IT infrastructure challenges that enterprise IT is not used to handling. However, most of the infrastructure requirements and challenges can be easily addressed with proper planning. In this guide, we’ll tell you how.

The first two layers of this four-stage architecture are in the realms of operational technology (OT).

Layer 1: Understand the epidermal layer of IoT — sensors and actuators

Sensors are devices that can detect and collect signals from the physical environment or the object that’s being measured and convert them into useful data.

For instance, the gyrometer and accelerometer in your smartphones are sensors that capture data from the physical environment and convert them into information that the phone can use to perform meaningful functions, such as changing the screen layout. Actuators can intervene and change the physical conditions that generate data, for instance, shutting off a power supply when temperature crosses a threshold.

The three forces that have made industrial sensors and actuators so powerful are:

  • Low-power wireless sensor networking technologies.
  • Power over Ethernet; enabling objects to operate without AC power supply as long as they’re on a wired LAN.
  • Miniature circuits that pack strong processing power on small surfaces.

Data processing happens at each stage of IoT architecture. While dealing with the first layer (that of sensors and actuators), IoT architects need to decide on tradeoffs among immediacy and depth of insight. The more immediate the need for data, the closer to the endpoints (sensors) the processing needs to occur. For applications that require advanced data processing, the sensor-generated data needs to be transmitted to centralized IT assets such as cloud platforms or datacenters.

So, for low-latency applications, such as course corrections of drones (the drone scene in “Transformers 5” was funny and exciting, and those drones were high tech and amazing) and autonomous cars (hopefully they are not hijacked like they were in “Fast 8”!), the data processing is preferably done close to the endpoint (IoT device).

Layer 2: Internet gateway

Sensors create a lot of data, and all of it is in analog form. Now, there are certain data management challenges here:

  • Analog data has specific structural and timing characteristics that need specialized software for processing; this can complicate the IoT architecture unless analog to digital conversion takes place before the data is sent to central IT assets.
  • Depending on the complexity of the IoT device, the number of sensors generating continuous data streams could be anything between two and 20. Even if it’s just two sensors operating, the volume of data is huge. Transmission of this massive volume of data to datacenters can pose tough demands on infrastructure resources.

To meet these challenges and make enterprise IoT successful, data acquisition systems (DAS) need to be used close to the sensors. These DAS perform data aggregation and conversion from analog to digital and feed the processed data to Internet gateways, which further route it through wired LANs, WiFi, and the Internet.

These systems sit in proximity to sensors and actuators; for instance, a jet engine that has sensors to measure velocity, temperature, pressure, vibrations, voltage, orientation, etc. could have a DAS attached within, accepting, aggregating, and converting data from all sensors.

The next two stages are in the realms of IT:

Layer 3: The edge

Aggregated and digitized data is sent over the Internet to datacenters. However, this data could well need more processing before it becomes viable for feeding to the datacenter. This is where edge computing systems come to the fore. Edge computing is beyond this guide’s scope, but TechGenix has done several excellent articles on the technology, which you can read here and here.

Edge IT units could be placed in remote offices. Now, because IoT device data can eat up bandwidth, edge computing resources are required to moderate the amount of data flowing to the datacenters. Instead of feeding all IoT data via one pipeline to the datacenter, it makes sense to perform some basic analytics and processing at the edge of the network.

For instance, when an industrial pump sends raw vibrations data over the Internet, edge systems can be used to crunch some numbers and only send plotted variations or predicted signal levels to the datacenter. Integrated computer systems like hyperconverged infrastructures are useful for edge applications because they’re fast and easy to remotely manage.

Layer 4: Central IT assets

These are your cloud-based data analytics engines, on-premises platforms, or hybrid setups, capable of performing advanced and deep data analytics to create astounding insight from data. These assets perform advanced data processing that is not governed by low-latency demands from the IoT application, such as prediction of when an industrial device is likely to malfunction.

These powerful IT systems also store and archive the humungous IoT data they receive, often with in-built mechanisms that help determine the usefulness of data for long-term storage. More comprehensive data analysis, based on data from disconnected endpoint devices, is also carried out at this stage. Regardless of the nature of the system (cloud-based, on-premises, hybrid), the nature of the data analysis carried out here remains the same.

What enterprise IoT architects need to be aware of

At the moment, the lines between these four layers are pretty visible; however, as layer-specific technology artifacts become more mature, there will be blurring of the boundaries, and that will require IoT architects (this has nothing to do with the dream sequence or fantasy world architects that we saw in the fantastic and profound movie “Inception”) to be aware and knowledgeable of all aspects of operations and information technology.

Photo credit: Shutterstock