Edge Computing: Processing Data Near the Machine

Why the Cloud Alone Is Not Enough

Imagine a bottling line in a beverage factory running at 1,000 bottles per minute. A smart camera inspects every bottle for defects. If every image is sent to a cloud server in another continent for analysis, the round-trip latency (50-200 milliseconds) means dozens of defective bottles pass before the result arrives. The solution is Edge Computing — processing data where it is generated, right next to the machine.

Edge computing moves processing and analytics capabilities from remote data centers to the closest possible point to the data source. In the industrial context, this means an industrial PC or smart gateway installed inside the control cabinet or next to the production line.

Cloud, Fog, and Edge: What Is the Difference?

To understand edge computing, it helps to see where it sits in the processing hierarchy:

Cloud Computing

Massive remote data centers (AWS, Azure, Google Cloud) provide virtually unlimited compute resources. Ideal for historical data storage, machine learning on large datasets, and management dashboards. However, they are physically distant and depend on internet connectivity.

Fog Computing

An intermediate layer between the cloud and field devices. Typically local servers in the control room or the plant data center. Offers lower latency than the cloud and continues operating when internet connectivity is lost.

Edge Computing

The closest point to the machine: a smart gateway, industrial PC, or even an advanced PLC. Provides near-instantaneous processing with sub-millisecond latency. Operates fully independently of any external connection.

Why Edge Computing in Industry?

Latency Reduction

In time-critical applications — safety systems, motion control, quality inspection — every millisecond matters. Edge computing eliminates the delay of sending data to a remote server and waiting for a response.

Reducing Data Volume

A modern production line can generate terabytes of data per day (cameras, vibration sensors, quality data). Transmitting all of this to the cloud is expensive and requires enormous bandwidth. Edge computing processes data locally and sends only results and summaries upstream.

Offline Operation

Factories in remote areas or in countries with unreliable internet cannot depend entirely on the cloud. Edge computing guarantees continued operation even when connectivity is completely lost.

Security and Privacy

Sensitive data (production recipes, quality data) stays inside the factory instead of traveling over the internet. This reduces the attack surface and complies with intellectual property protection requirements.

Industrial Edge Gateways

The edge gateway is the core device in industrial edge computing. It combines capabilities for:

• Data collection: from PLCs via OPC UA, Modbus, or PROFINET
• Local processing: running analytics algorithms and fault detection
• Store-and-forward: buffering data during connectivity loss and transmitting it later
• Cloud connectivity: sending processed data to the cloud platform via MQTT or AMQP

Examples from the market:

Containers at the Edge: Docker and Kubernetes

Why Docker in the Factory?

Docker packages an application with all its dependencies (libraries, configuration) into an isolated container. The benefits in an industrial setting:

• Portability: the same container runs on any Docker-capable device without modification
• Isolation: multiple applications run on the same device without conflicts
• Seamless updates: update one container without affecting the others
• Fast rollback: if an update fails, revert to the previous version in seconds

Example: an edge gateway running 4 Docker containers:

1. Data collection container: reads from PLCs via OPC UA
2. Analytics container: runs a machine learning fault detection model
3. MQTT broker container: publishes data to subscribers
4. Local dashboard container: a web interface displaying machine status

Kubernetes at the Edge (K3s / MicroK8s)

When you have dozens or hundreds of edge devices, you need a centralized management system. K3s (a lightweight Kubernetes distribution) enables:

• Deploying applications to all devices from a single point
• Rolling updates with zero downtime
• Automatic container health monitoring and restart
• Resource management (CPU, memory) per container

Industrial Edge Platforms

Major manufacturers offer integrated platforms that combine hardware and software:

Siemens Industrial Edge

• Integrates with existing SIMATIC systems
• App store for ready-made edge applications
• Supports Docker and Node-RED alongside custom applications
• Centralized management of all edge devices via Industrial Edge Management

AWS IoT Greengrass

• Extends AWS Lambda to the edge
• Supports local machine learning inference
• Automatic cloud synchronization when connectivity is available
• Runs on resource-constrained Linux devices

Azure IoT Edge

• Runs Azure Modules locally
• Supports Azure Stream Analytics at the edge
• Integration with Azure Digital Twins
• Device management via Azure IoT Hub

Practical Example: Edge-Based Quality Inspection

Here is a design for a quality inspection system on a packaging line in a food factory:

Requirements:

• Inspect 500 packages per minute
• Detect defects (tilted label, open cap, misprinted date)
• Response time under 50 milliseconds to reject a defective package

Architecture:

1. Industrial camera (e.g., Basler ace) captures an image of every package
2. Edge computer with a GPU runs a computer vision model
3. Digital output triggers the pneumatic rejector
4. Cloud connection sends daily quality statistics and defect images for analysis

All processing happens locally on the edge device. The cloud is used only for historical data storage and periodic model retraining.

Edge Computing Challenges

Despite its advantages, edge computing comes with challenges:

• Managing distributed devices: hundreds of scattered devices are harder to manage than a single cloud server
• Security: every edge device is a potential attack point. It must be secured with TLS certificates, security patches, and access restrictions
• Limited resources: processing and storage capacity is constrained compared to the cloud
• Data synchronization: when connectivity returns after an outage, data must be synchronized without duplication or loss
• Software lifecycle: updating applications across hundreds of devices requires automation (DevOps/GitOps)

Summary

Edge computing is not a replacement for the cloud — it is a complement. Data that requires immediate processing is handled at the edge, while data that needs deep analysis and long-term storage goes to the cloud. With container technologies (Docker/K3s) and industrial edge platforms, factories can benefit from the best of both worlds: edge speed and cloud power. For automation engineers, edge computing provides a practical solution that combines real-time responsiveness with advanced analytics capabilities.