What Is Edge Computing?
A practical introduction to processing data close to physical devices, people, and operations.
Edge computing places data acquisition, processing, storage, and operational response near the point where data is created. It does not eliminate cloud or central systems. It gives physical operations a local boundary that can continue to work, retain context, and deliver selected outcomes when a remote connection is slow, expensive, unavailable, or simply unnecessary for the first decision.
Move the first useful decision closer to the source
In a cloud-only design, a device sends data to a remote service before an application can analyse it or make it available to an operator. In an edge design, a local computer or gateway near the device handles the work that benefits from proximity: connecting equipment, validating data, buffering inputs, running a local task or model, retaining evidence, and creating an event or response.
The edge can be an industrial PC, server, compact Linux host, on-board computer, field gateway, or another owner-controlled device. Its location is defined by the operational boundary, not by a distance from a data centre. A plant room, vehicle, store, farm, depot, cabinet, laboratory, or remote facility can each be an edge site when it owns a distinct physical process and local failure behaviour.
Keep raw inputs, local work, and shared outcomes distinct
| Layer | Role | Examples |
|---|---|---|
| Physical sources | Produce measurements, media, state, or control signals. | Cameras, sensors, PLCs, meters, gateways, scanners, robots, and local services. |
| Edge node | Connect, timestamp, validate, retain, and process inputs close to the source. | Industrial PC, local server, ARM device, vehicle computer, or site gateway. |
| Local operating logic | Turn source state into a bounded, explainable local result. | Freshness check, threshold, local model result, event rule, evidence retention, or operator view. |
| Central or external consumers | Use the selected data package for coordination, reporting, history, planning, or further analysis. | Cloud applications, dashboards, WMS, CMMS, SCADA, data lake, maintenance system, or partner API. |
Edge and cloud are usually complementary
| Keep work at the edge when | Use a central or cloud service when |
|---|---|
| Raw data is high-volume, time-sensitive, confidential, or still operationally useful during an upstream outage. | Many sites need long-term aggregation, portfolio reporting, cross-site planning, large-scale training, or shared business workflows. |
| A local asset needs a result quickly enough to support an operator, bounded rule, or local evidence workflow. | The data can tolerate transfer delay and benefits from central scale, collaboration, or compute capacity. |
| The owner must keep raw recordings or measurements within a specific local boundary. | Approved derived events, aggregates, or data packages need to reach other authorised consumers. |
The useful question is not “edge or cloud?” It is: which data and decisions must remain local, which outcomes should be shared, and what should happen when the connection between them is unavailable?
Edge computing is an operating model, not only a hardware location
An edge node needs more than processor capacity. It needs a defined device contract, local storage and cleanup rules, source freshness checks, health monitoring, access control, update procedures, and a clear handoff to external consumers. Without these, a local computer becomes an unobservable integration point rather than an operational boundary.
Banalytics illustrates this model through a local Agent that connects physical resources, manages components and tasks, creates events and actions, retains configured data, and exposes selected outputs to dashboards and integrations. Its System Monitor makes CPU, memory, disk, thread, media, and connection state available as operational events; Portal Integration supports authorised remote browser access without putting the portal in the local acquisition path.
Choose edge computing when proximity changes the outcome
- Local data or evidence remains useful when the WAN is unavailable.
- The first response has a time, privacy, bandwidth, or reliability requirement that a cloud-first path cannot meet.
- Devices and legacy systems need a local integration boundary before they can expose selected outcomes.
- Teams need to preserve the authority of local safety, control, or system-of-record platforms while improving visibility around them.
- The organisation can name the owner of the site, data, operational response, and lifecycle.
Edge computing is not a requirement to move every workload out of the cloud. It is a way to give physical operations the local context and resilience they need, then connect them to wider systems on deliberate terms.