What “Edge” Actually Means
Microsoft’s cloud computing dictionary defines edge computing as a framework that “allows devices in remote locations to process data at the ‘edge’ of the network, either by the device or a local server. And when data needs to be processed in the central datacenter, only the most important data is transmitted, thereby minimizing latency.”
That definition is a starting point, but three practitioners in the space—Jerome Hardaway, senior software engineer at Microsoft; Kate Goldenring, co-chair of the CNCF IoT Edge Working Group; and Alex Ellis, founder of OpenFaaS—each suggest it needs more nuance. Hardaway would make it human-centric: edge computing is about processing data as close as possible to the person using it, since a device without a user is just hardware. Goldenring points out that definitions generally fall into three categories: geographic (distance between devices and servers), resource-based (constraints on compute and storage), and connectivity-based. Ellis notes the term is broad enough that two experienced engineers can mean entirely different things when they say “edge.”
Edge vs. Traditional Architectures
The line between edge computing and private or hybrid cloud can blur, since all involve some on-premises compute. The differentiator is intent. Hardaway says private and hybrid cloud is primarily about controlling where data resides—a healthcare provider keeping patient data on-premises, for instance—whereas edge computing targets specific requirements like low latency or reliable operation under constrained conditions.
Goldenring frames edge as a continuum: downstream devices, upstream cloud resources (public or private), and intermediate nodes. You have to consider what storage and compute are available at each point, especially given that network connectivity is often a major constraint at the edges. Ellis adds that resource constraints aren’t universal—some edge devices are quite capable—but you must design for environments where assumptions like a persistent TCP/IP connection don’t hold. A battery-powered remote sensor might only connect intermittently to conserve power, and may need to communicate over a low-power wireless protocol to an intermediary device.
When Edge Doesn’t Fit
Edge computing adds intermediaries between a device and a data store, which expands the attack surface you have to secure. Hardaway warns that industries like healthcare must weigh those trade-offs carefully and understand the business problems they’re solving before committing to an architecture.
Ellis cautions against pigeonholing edge use cases. Building edge systems for medical devices differs fundamentally from doing so for restaurant point-of-sale terminals or airlines. A POS device may run full Android and tolerate periodic reboots for updates; a medical device likely needs a real-time OS that runs continuously. Goldenring says the deciding factor is application characteristics: highly stateful apps that need constant connectivity and lots of storage are better off in the cloud, though intermediate nodes near the edge can still offload lighter processing.
Portability Across Platforms
Edge applications tend to be deeply customized to their circumstances, which limits portability. Goldenring notes that the many ways to connect to devices require custom logic, but you can standardize what happens after data is ingested. Hardaway adds that the more you leverage a platform’s built-in features, the harder it becomes to move, though implementing your own solutions means more work to maintain. That dependency isn’t inherently bad, but it must be a conscious choice. Ellis points out that if your hardware can run a container or VM, your solution can be quite portable; bespoke embedded software is another story.
Skills for Edge Development
What you need to learn depends on the type of edge work you’re pursuing. Ellis’s Linux Foundation course on Kubernetes and K3s at the edge highlights useful adjacent skills: MQTT, shell scripting, and Linux. Kubernetes itself is broadly useful, he says, whether you’re making a web app more responsive by moving resources closer to users or working with industrial equipment.
Hardaway recommends Python as a language that interoperates across platforms and environments, calling it one of the most transferable technical skills. He also suggests professional certifications in edge technologies—one of the few areas where he recommends them—because they signal genuine investment in learning. Communication skills matter too, he adds, because mistakes are inevitable.
Goldenring suggests starting with networking fundamentals and the protocols devices use to communicate. Beyond that, she recommends hands-on experimentation: get a Raspberry Pi, connect it to an edge computing platform, and learn the terminology by doing. “Have some fun, that’s the best way to learn.”
