API Management / Edge / IoT / Machine Learning / Contributed

The Internet of Things Requires a Connected Data Infrastructure

2 Sep 2021 8:55am, by

Danny Bedgood
Danny Bedgood is the worldwide VP of Industry Vertical Solutions at Aerospike. He joined Aerospike after a 20-year career at Hewlett Packard Enterprise. At HPE, he pioneered efforts in IoT, Edge Computing, Artificial Intelligence and Machine Learning, as well as automotive market transformation initiatives related to mobility and advanced driver assistance systems (ADAS).

Mobility is one of the hottest topics for the Internet of Things (IoT) today. A few years ago, use cases often centered on connected devices in the home or business. But in 2021, IoT is on the move. Cars are a perfect example.

Vehicles these days are much more than an engine, drivetrain and four wheels. Today, cars and trucks should be thought of as mobility platforms. They gather data from all kinds of sources — from phone devices to navigation, engine, steering — even seats (carmakers are putting in biometric seats that will record health information). But this is merely the tip of the iceberg.

Great strides are being made toward the practical use of autonomous, or self-driving, cars. As these vehicles become mainstream, it will represent a 100-year paradigm shift for the industry. If humans aren’t at the wheel, then what drives these vehicles? Data.

Databases and edge data processing capabilities will be the linchpins for success in all IoT projects. The more data that is ingested and the faster it’s processed — the better will be the decisions. Using the maximum number of data sources creates a better customer experience and more efficient business processes.

Current “connected” vehicles and future autonomous machines consist of a myriad of sensors that generate a huge amount of data, and this data is often processed at different speeds. Some of the data is structured; other data is not. A dizzying array of data coming at different velocities for different purposes poses a massive data challenge.

Not long ago, a terabyte of information was an enormous amount and might be the foundation for solid decision-making. These days, it won’t cut it. For example, looking at a terabyte of data might yield a decision that’s 70% accurate. But leaving 30% to chance is unacceptable when it comes to real-time vehicle safety. On the other hand, having the ability to ingest and process 40 terabytes — from all sources, edge to core — can result in an accuracy rate well exceeding 90% accuracy. Something jumps in front of your car — is it a person, a dog, a trash bag, a child’s ball? Real-time systems need to determine the level of risk and react in micro milliseconds.

Real-time processing has to be done closer to where the decisions are being made. In terms of IoT, a lot of questions can be answered by using a digital twin. These create additional layers of insights and provide a better understanding of what’s happening in any given situation and decide on the most appropriate course of immediate action.

Digital twins take insight not just from the raw sensors — the edge compute nodes — but a combination of real-time data at the edge and historical data at the core. Separate systems should not be used to manage the various data sources and fuse them together. That’s too complex and time-consuming. Apache Spark is ideal for fusing data and creating layers that enable us to create new rules and machine learning models. Priceless insights flow from these models.

Data is the future for IoT initiatives in the automotive industry as well as many other markets. In order for IoT processes to flourish, companies need a connected data infrastructure.

Transforming IoT data Architectures

Traditionally, IT architectures have been focused on the core environment. Applications operating at the edge of an enterprise have been required to pull data from the core. But that’s not fast enough anymore. Fortunately, processing has progressed to the point where newer, smarter technology can operate on the edge. This includes hardware and storage, which occupy much less of a footprint than in the past.

Transforming the IT architecture makes it possible to tie into both data at the edge and the core system. With processing now taking place at the edge, a whole stack of latency and communication issues are relieved. In fact, things can be run autonomously. Even if there’s a disconnect, a system can still manage,  perform and drive decisions at the edge without the core system. This represents a significant shift in the thinking of architects around the world.

Real-time monitoring capability is especially vital with IoT. It enables organizations to see what’s going on and engage with their systems. Companies can know what’s happening at any given moment, regardless of location. There’s no reason to wait for static reports anymore. Organizations can now monitor performance and quality issues by pulling everything into a screen instantaneously. For the auto industry, this can be information from production facilities to showrooms to drivers on the road.

Comprehensive Edge-to-Core IoT Blueprints

Having a common set of architectures and building them into a repeatable solution is a must. Look at today’s global supply chain. The slightest disruption can be absolutely crippling for companies. The auto industry has been a prime example in 2021. The global shortage in computer chips has had a ripple effect throughout the market. Production facilities have been stalled, and a restricted supply of vehicles has led to price increases for consumers.

Companies must take an unflinching look at their own operations. A real-time architecture is a must. There are stacks for hardware, the network, communications and software. Repeatable, full-blown blueprints are needed that illustrate how things are loaded, how things are implemented and how consoles are built. Having a real-time architecture in place will provide an ability to react to the vagaries that confront enterprises every day.

The Role of AI and ML Analysis to Speed and Optimize IoT Data

Artificial Intelligence (AI) and Machine Learning (ML) have become increasingly important — they have also become far more complicated and demanding for systems. Yesterday’s simplistic algorithms with a handful of tasks have been replaced by algorithms with thousands of requests. This is an enormous shift and it requires systems that have the ability to process larger, more complex AI/ML strings while retaining real-time decisioning. That represents a dramatic new shift.

For enterprise-level companies, pet-scale data environments that include real-time decisioning are table stakes. Companies are realizing that having the ability to write really complex AI/ML strings and load those into decisioning engines that don’t degrade the system is a huge asset. In some instances, it’s even possible to do it on less hardware. Having a smaller architecture footprint is typically more efficient (less Capex and Opex, best of both worlds). Yet, users can still go as fast and deep as they need to — with the ability to grow as needed.

Having AI/ML-powered real-time infrastructures in place, and the blueprints to spawn others as needed, will enable an unprecedented level of decision making. If, as many contend, information is the new currency, then the real-time database platform is the “reserve system” of the enterprise — underpinning and guaranteeing greater levels of success.

The New Stack is a wholly owned subsidiary of Insight Partners. TNS owner Insight Partners is an investor in the following companies: MADE, Real.

Feature image via Pixabay.

A newsletter digest of the week’s most important stories & analyses.