The edge, or edge computing, is messy. Defining what it is, where it exists, and what it does is very much determined by the perspective of those using or selling it. 

As such, when a group of executives from cloud and edge computing vendors were asked to define the edge during last week’s Kubecon and CloudNativeCon event, the responses were not uniform. 

“Our definition of edge is the network edge,” said Nancy Li, director of edge cloud platforms at Cox Communications. Device manufacturers think the edge is in the device and “when you talk to data centers, they think the edge is directly in those data centers — colocations,” she said. 

“We think the edge is more towards the last mile of the network where we can introduce compute, storage, and [artificial intelligence] services running on top of our edge,” Li explained.

Jumping off those comments, Dejan Bosanac, senior software engineer at Red Hat, noted that the edge is defined by the interests and current perspectives of whoever is being asked to clarify its role. “In my opinion it all boils down to extending the cloud, so extending the cloud-native development practices is what basically defines edge computing,” he said. 

There are multiple tasks or use cases within that framework, including the provision of proper infrastructure and development practices to enable edge computing, “but it’s all about extending the cloud platforms,” Bosanac explained. 

Michael Michael, director of product management at VMware, offered the most detailed definition of edge computing by framing it around use cases and capabilities. “For different industry verticals the edge looks a little bit different,” he said, adding that it helps to view the edge based on the workloads that can be run in a cloud-native environment. 

The edge, he said, can be classified into three categories: thick, medium, and thin. The “thick edge” includes enough compute power to accomplish many of the same tasks that are typically run in data centers. “Think of it almost like a regional, small-scale data center,” Michael said. 

A “medium edge” environment includes a couple racks of servers that can run data-tier applications and cloud-native workloads. And finally, a “thin edge” facilitates the delivery of services for 5G, smart cars, IoT devices, and small businesses that may only have the capacity for a small compute node with less strict requirements for memory and CPU, he explained. 

“So anything outside the core data center for us is edge, but we classify into these areas that kind of identify what kind of workloads can you run there,” Michael said.

The definition of course, still unravels further from there, based on staffing, service needs, and local or remote management requirements. There are also inherent restrictions and resource constraints in edge computing, including latency, processing, bandwidth, and connectivity, he said.

“Edge is not a new concept. … It’s very heterogeneous and diversified,” said Cindy Xing, principal software engineering manager at Microsoft.

“I would like the industry to come to some consensus about the terminology,” she said. “In my mind, edge is differentiated from cloud. So it’s where the intelligence happens close to where the data is originally generated.”

LF Edge, an umbrella organization under the Linux Foundation, recently updated the taxonomy and framework of edge computing in a report but even its attempts to coalesce around a more streamlined definition led to more caveats and complexity. 

“We typically think of the edge or the edges in terms of the physical devices or infrastructure where application elements run,” the organization wrote. “However, the idea of a path between the centralized cloud and the device edge instead allows for the conceptualization of multiple steps along the way.”

LF Edge landed on two main categories — the service provider edge and the user edge — but broke those down even further into subcategories. These efforts paired with the discussion at last week’s event further illustrate the fact that as long as the edge remains one of the most hyped areas of development in cloud computing and networking, what it is, what it does, and where it resides will remain — well, cloudy.