Common Use Cases and Examples of Edge Computing

Edge computing covers a huge number of industries, implementations, and use cases. Finding examples of edge computing in the real world today isn’t hard — what’s more challenging is understanding whether a given example meets the definition of edge computing, which always seems to be shifting.

The best way to consider the “edge or not” question is often to evaluate where the inertia of computing lies in a given use case: Is it becoming more centralized, or is it becoming more distributed? If it’s the latter, you’re likely looking at an edge computing scenario. The edge, as a reminder, is the space at the furthest reaches of a given piece of digital infrastructure. But this definition, too, is very fuzzy — many edge computing definitions say that end devices are beyond the edge, and others contend that end devices may be edge computing devices themselves, based on the particular context. Suffice it to say, there is no definitive answer for what constitutes the edge or what precisely defines an edge device.

What are Examples of Edge Computing?

• Content delivery networks: CDNs are perhaps the biggest and oldest example of edge computing. By placing servers strategically in locations where demand is anticipated, a service can deliver content (video, cloud data, web pages) much more quickly — and with greatly reduced burden on overall web infrastructure. CDNs are a massive industry, and their entire premise is built upon placing resources at the edge.
• Edge services: What do video game streaming and AI have in common? They’re both use cases that benefit greatly from the edge services model. By placing the computing resources that deliver these services closer to customers, the quality of these services can be vastly improved — lowering latency, increasing QoS, and reducing data use.
• Vehicle-to-everything (V2X): Whether it’s V2V (vehicle-to-vehicle), V2I (vehicle-to-infrastructure), or V2G (vehicle-to-grid), or autonomous vehicles (AV), modern vehicles are increasingly equipped with powerful onboard computers and sensors. Processing camera and lidar data from the road in real time for automation, plugging in to the local power grid to act as a “virtual power plant,” and interfacing directly with our smart devices (phones) to provide enhanced services, cars have become an edge category unto themselves.
• Manufacturing and industrial automation: With advanced sensors and robust machine-learning assisted computing, manufacturing and industry have become hubs of edge innovation. A factory that can see products coming off the line and then apply computer vision algorithms locally to determine if output is meeting quality and performance metrics is a game-changer, and sophistication and automation in industrial edge computing are constantly increasing.
• Healthcare: More and more medical devices are gaining connectivity capabilities, processing data locally and reporting results directly to providers on-site. This allows medical professionals to make decisions more quickly and with better patient data, and for patients to receive higher quality care. 
• Retail automation: Fully-integrated self-checkout and point of sale systems with payment processing, inventory reporting, video surveillance, and printing are increasingly commonplace in retail. Retail is rapidly embracing automation and edge computing to decrease reliance on manual processes, respond to demand with more agility, and increase operational efficiency.
• Edge IoT: The use of dedicated sensors that can collate and analyze data locally is growing in almost every space. Whether it’s scientific sensors for remote data gathering, industrial sensors for remote oil and gas operations, or building sensors (e.g., doors, HVAC, occupancy) for commercial property management, edge IoT sensors are proliferating at the edge at a breakneck pace.

What are Examples of Non-Edge Computing?

One of the easiest ways to understand what meets the definition of edge computing is to understand what doesn’t. Here are some examples of computing that would not be considered “edge” — and counterexamples that are.

• Traditional POS retail: If your retail stores report back raw data about sales and inventory to a database running on a central server for analysis and storage, this is a classic example of non-edge computing. 
  • The edge version of this use case would store, process, and optimize sales and inventory data locally before reporting it to a central data repository. This could speed up data insights, allow more agile inventorying, and reduce supply chain lag.
• Traditional web portal services: If all of your users (for example, tellers at a bank doing data entry or accessing customer account information) access a given resource over the web via a secure connection to a centrally administered and controlled server, this is not edge computing.
  • An edge version of this use case might use an edge server where the web service runs locally, caching data relevant to a specific location or region. This could greatly increase the speed of data entry, by processing requests in the background instead of requiring the user to wait for a response from the central server each time data is sent. Local customer and branch data could be cached for quicker access. Business conducted between local branches could occur on the edge server, rather than being reconciled centrally.
• Traditional financial tech devices: If your service uses an intermediary device to send a request to a central processor for data reporting and authentication — for example, a handheld credit card processor or an ATM — this is not edge computing.
  • An edge version of this use case would see devices reconciling and authenticating transactions locally, perhaps using P2P with a customer’s end device as a trust layer. This could allow payment processing without always-on connectivity, or even true peer-to-peer payments without a processing intermediary.
• Mainframe-and-terminal computing: Edge computing requires substantial computing to occur away from a central computing resource. Fully closed systems like traditional mainframe-and-terminal systems (still operated by some airlines) are a great example of a non-edge architecture. All terminals must remain connected to the mainframe system to remain operational, and every request must be processed on that mainframe — terminals act only as interfaces.
  • This is about as “non-edge” as it gets. There’s no edge version of using a mainframe!

Where is Edge Computing Headed in the Future?

Edge computing trends tend to be visible in imagined futures like smart cities, vehicle-to-everything connectivity, connected infrastructure, and other distributed computing models. Much in the way electric car visionaries espouse a narrative of a time when every vehicle connects to the grid to create an always-on, ultra-resilient, highly distributed supply of electricity with no single point of failure, so too do edge computing optimists talk about the future at the edge.

In a world where cars, electrical grids, municipal services, traffic infrastructure, and edge AI servers running on software-defined networks are constantly exchanging information, much of the need for highly centralized computing resources will disappear. Instead, computing will scale in a way that’s inherently proportional to the density and demand of users. For example, a large gathering of people inside a stadium will trigger algorithms to intelligently reroute traffic for the best flow, increase public transit frequency to the stadium, alert service providers of a trending high-demand event, and dedicate network resources to guarantee quality of service (QoS) to all relevant endpoints in the area. The edge, in an ideal world, is a highly responsive, adaptive ecosystem of computing resources working in concert to seamlessly deliver the best experience and outcome possible.

Of course, that’s pretty pie in the sky stuff. Today, the edge is mostly about getting your Netflix delivered in 4K with minimal buffering and keeping a close eye on critical industrial systems.