Edge Computing

What is Edge Computing?

Edge Computing is an architecture in which data is processed locally at the edge of the network, close to the source such as a sensor, machine or PLC, instead of in a central data centre or the cloud. Only the result, for example a summary, an alarm or a prediction, is then passed on to central systems. In industrial automation, edge computing forms the bridge between the OT world of machines and controllers and the IT world of analytics, ERP and the cloud.

Edge Computing = processing data where it is created.

The goal is to react faster, use less bandwidth and depend less on an internet connection. In 2018 Gartner estimated that around 10% of enterprise data was created and processed outside a data centre or cloud, and predicted roughly 75% by 2025. Factories, power grids and water utilities are exactly the environments where that data is generated.

🧠 What does edge computing do in a factory?

In practice an edge solution performs four tasks:

• Collect - reads data from PLCs, SCADA, HMIs and sensors via protocols such as Modbus, OPC UA or PROFINET
• Pre-process - filters, compresses and aggregates raw measurements, for example from thousands of vibration samples per second down to one characteristic value per second
• Analyse - runs rules, machine learning models or image recognition locally and flags anomalies
• Forward - publishes results to a historian, MES or cloud platform, usually via MQTT, and buffers data when the connection drops (store-and-forward)

The edge does not take over control tasks. The PLC controls the process deterministically with cycle times of a few milliseconds. The edge observes, computes and advises; it typically works in soft real time, not hard real time. Safety functions such as an emergency stop or a process shutdown always belong in the PLC or a safety PLC, never on an edge device.

🔄 What is the difference between edge, fog and cloud?

The terms are often used interchangeably, but in 2018 NIST drew a clear distinction in SP 500-325 (Fog Computing Conceptual Model). There, fog is an intermediate layer between end devices and the cloud. The OpenFog reference architecture was adopted in 2018 as IEEE 1934-2018, and the OpenFog Consortium merged into the Industrial Internet Consortium in January 2019.

In practice vendors usually speak only of edge and cloud, with fog falling under “site-level edge”. The latency figures are indicative and depend on the network and distance.

🔧 What edge hardware is available?

Modern platforms almost always deliver software as Docker containers. This lets you roll out, update and roll back apps centrally, just as in the IT world.

🏭 What are the use cases of edge computing in OT?

• Predictive maintenance - a vibration sensor samples at tens of thousands of samples per second; the edge calculates RMS and peak values and forwards only those characteristics and alarms
• Quality control - an AI model analyses camera images on the line and rejects products within milliseconds, without sending images to the cloud
• Protocol conversion - a gateway translates legacy protocols such as Modbus or S7 into OPC UA or MQTT, so existing machines can connect to modern systems
• Remote sites - a wind turbine, pumping station or drilling platform processes data locally and sends only relevant information over a narrow link
• OEE and energy - local calculation of availability, performance and energy consumption per machine for shop-floor monitoring

🧱 How does edge computing fit into a Unified Namespace?

In a Unified Namespace (UNS) every data source publishes its current state to one central MQTT broker, organised by site, area, line and machine. Edge devices are the natural publishers: they read the PLC, add context (unit, machine, status) and publish to the right topic. With MQTT Sparkplug B this happens with a fixed payload structure and reporting of online and offline state.

This makes the edge the starting point of Industrial DataOps: data gets its meaning at the source rather than afterwards in a data warehouse. Low-code tools such as Node-RED and platforms such as Ignition are widely used for this translation step.

☁️ Which edge platforms are there?

There are also vendor-neutral platforms from SCADA and DataOps vendors, among others. When choosing, pay attention to support periods: an edge platform often has a shorter life than the machine beneath it.

🔐 How do you secure edge devices?

An edge device connects OT with IT and often with the cloud, which makes it an attractive target. Local processing can improve cybersecurity because less data leaves the site, but only if the device itself is properly secured. Edge security therefore needs the same discipline as any other system on the boundary between IT and OT.

• Placement in the Purdue Model - edge devices belong in their own zone, usually at level 2 or 3, with the cloud connection running via the IDMZ; prefer outbound connections (MQTT over TLS) so no inbound ports need to be opened
• Hardening - disable unused services and ports, use secure boot and a TPM for key storage
• Identity - give every device its own certificate instead of shared passwords, handled through certificate management
• Updates - organise patch management for the operating system, container runtime and apps, with the ability to roll back
• Standards - IEC 62443-4-2 sets requirements for components such as embedded and host devices; network segmentation limits the damage of a breach

For vendors, the Cyber Resilience Act applies. Since 11 September 2026 manufacturers must report actively exploited vulnerabilities and severe incidents, with an early warning within 24 hours. From 11 December 2027 all requirements apply, including a support period for security updates of, in principle, at least five years. Users covered by NIS2 must include edge devices in their risk assessment and security monitoring.

🧭 How do you get started with edge computing?

1. Choose one use case - for example pump monitoring or OEE on one line, with a measurable goal.
2. Inventory data sources - which PLCs, protocols and tags exist, and who owns them?
3. Decide where processing happens - what must stay local (speed, availability, confidentiality) and what can go to the cloud (training, reporting)?
4. Select hardware and platform - check industrial temperature ranges, fanless design and the support period.
5. Design zones and connections - define how the device joins the network, which direction traffic may flow and how remote management works.
6. Arrange management - central rollout, updates, backups and monitoring, with clear agreements between IT and OT.
7. Scale up - roll the pattern out to more lines, preferably with a shared data model or Unified Namespace.

❓ Frequently asked questions

Is edge computing the same as IoT?

No, edge computing is a way of processing data, whereas IoT describes connecting devices to networks. Edge computing is often used within IIoT architectures to process data from connected devices locally. An IoT solution can also work without an edge by sending all data straight to the cloud.

Does edge computing replace the PLC?

Edge computing does not replace the PLC. The PLC performs deterministic control and safety functions, while an edge device collects, analyses and forwards data. Virtual and soft PLCs that run on edge hardware do exist, but they require extra attention to real-time behaviour and certification.

What is the difference between an edge device and a gateway?

A gateway mainly translates protocols and forwards data. An edge device can also compute, run apps or containers and perform local analytics. In practice the terms overlap, because many modern gateways can run containers too.

Does edge computing work without an internet connection?

Edge computing is designed precisely to keep working without an internet connection. The edge device continues processing data and buffers results until the connection is restored (store-and-forward). Only cloud functions, such as central dashboards or model training, are temporarily unavailable.

Is edge computing more secure than the cloud?

Edge computing is not automatically more secure than the cloud. Sensitive process data stays local, but every edge device is an extra system that must be secured, patched and monitored. With hardening, certificates, segmentation and outbound-only connections, edge computing can reduce risk.

Is edge hardware covered by the Cyber Resilience Act?

Edge devices and edge software placed on the EU market are products with digital elements and therefore fall under the Cyber Resilience Act. The reporting obligation for actively exploited vulnerabilities has applied since 11 September 2026, and the full requirements apply from 11 December 2027. Manufacturers must, among other things, provide security updates during a support period of, in principle, at least five years.

📌 In summary

Edge Computing brings compute power to where data is created, so factories can work faster, more efficiently and less dependent on the cloud. Edge devices collect and enrich OT data, run local analytics and publish results to a Unified Namespace or the cloud. Control and safety stay in the PLC, and the edge device itself must be secured as seriously as any other system between IT and OT.