Image: iStock/chombosan

According to a ForresterNow report, “The Four Edges of Edge Computing,” there are four types of edge computing but there are no hard and fast divisions between them. Also, the way vendors are using the term makes it hard to discern the subtle differences.

“We, as analysts, like to draw clear boundaries but, in truth, real life is much messier,” said Brian Hopkins, Forrester’s vice president of Emerging Tech Portfolio and report’s author. “It’s best to think about the edges not as strictly physical division of a network but as broad categories of related solutions that have things in common and therefore have attracted different vendors. The four edges is a response to clients whom I always have to explain edge to because different vendors use the word to mean different things.”

SEE: Edge computing adoption to increase through 2026; organizations cautious about adding 5G to the mix (TechRepublic Premium)

Hopkins said that, for example, Equinix, a co-location vendor, uses the word edge to describe running distributed enterprise workloads at various colocation data centers. For its part, internet of things (IoT) vendor Siemens uses the word edge to mean running software and analytics on servers that are located in industrial manufacturing spaces.

“These two scenarios are very different and are being serviced by different vendor solutions, even though they both use the term ‘edge’,” said Hopkins.

The report defines the edge as “infrastructure and software that is physically separate from but connected to enterprise core IT assets by shared fiber or wireless networks.” There are four types of edge computing outlined in the report; each serving specific niches: enterprise, operations, engagement and provider.

The enterprise edge is made up of second- and third-tier data centers, cloud gateways, co-location facilities, and office spaces connected by wide area networks (WANs). The goal of the enterprise edge is to help customers extend their core compute networks through enterprise infrastructure and WAN and public cloud vendors. Typical use cases include office automation, smart buildings and e-commerce.

The operations edge consists of IoT networks and devices, local compute assets and gateways. These edge networks help automate processes on-site at extended locations such as retail stores, branch offices, manufacturing facilities, warehouses, storage facilities, processing plants and the like. Typical use cases include industrial automation, smart city functionality and services, logistics and enabling experience spaces.

The engagement edge is defined by globally distributed, high-performance compute clusters designed to improve customer engagement through faster response times. These points-of-presence (POPs) networks connect customers to data and information faster by positioning it physically closer to the point of consumption. Typical use cases include streaming, gaming, zero-trust security and smart home functionality and services.

The provider edge is made up of telco-owned infrastructure and software. These “edge clouds” are designed by telcos to offer infrastructure-as-a-service and 5G enablement closer to the end-users but still within the operator’s networks. Customers use the telco’s compute and storage resources to automate software that supports the other three edges. Eventually, these edge clouds will be able to host operations and engagement edge functionality and data. Typical use cases include edge clouds and private 5G cellular networks.

SEE: AWS Lambda, a serverless computing framework: A cheat sheet (free PDF) (TechRepublic)

Even though edges are geared for different use cases, in the end, they just provide compute capability so they can be used for any number of scenarios. This causes confusion, as well, said Hopkins. A piece of code can run on different edges depending on the need. For example, some firms run IoT-type workloads, such as data processing for equipment predictive analytics, on engagement edge compute clusters rather than on edge servers located on the shop floor, said Hopkins. This represents an overlap between engagement edge and operations edge.

The second overlap happens when underlying infrastructure vendors sell compute, storage, networking and software that runs on many edges. For example, HPE makes servers that can run in the operations edge, the provider edge or the enterprise edge.

And, finally, some vendors offer edge-as-a-service. They own the entire compute stack so customers pay based solely on the services they are buying.

“Edge-as-a-service is also a marketing term that different vendors will realize differently,” said Hopkins. “As such, I suggest that clients ignore it and get to the reality of the solution being offered: What is the problem it solves? What is the set of hardware, network and software assets needed?”