Not every workload belongs in a central data centre or public cloud region. Edge computing places compute and storage close to where data is generated — in factories, stores, hospitals, vehicles or branch offices.
Why move compute to the edge?
- Latency: machine control, video analytics and point-of-sale systems need fast local responses.
- Bandwidth: processing video or sensor data locally avoids sending everything to the cloud.
- Resilience: sites can keep operating when connectivity is lost.
- Data residency: some data must stay on site for regulatory or contractual reasons.
Common enterprise use cases
Typical examples include quality inspection with computer vision in manufacturing, in-store analytics and inventory in retail, patient monitoring in healthcare and predictive maintenance for remote equipment.
Architecture choices
Edge deployments range from small ruggedised devices to compact server clusters. Many organisations use lightweight Kubernetes distributions or managed edge platforms so applications can be deployed and updated consistently across hundreds of sites.
Management challenges
- Remote, zero-touch provisioning for sites without IT staff.
- Centralised monitoring, patching and configuration.
- Physical security for devices in accessible locations.
- Secure connectivity back to the core, often through SD-WAN or SASE.
5 best practices for enterprise edge deployments
- Standardise hardware and software stacks. Identical, pre-configured edge nodes are far easier to deploy, support and replace than bespoke builds at each site.
- Use zero-touch provisioning. Devices should configure themselves securely when connected, without on-site IT staff.
- Manage centrally. Use a central control plane to deploy applications, push updates and monitor health across hundreds or thousands of locations.
- Design for disconnection. Edge applications should keep operating when the connection to the cloud or data centre is lost and synchronise when it returns.
- Secure the physical layer. Edge devices are often in accessible locations, so use tamper detection, encrypted storage, secure boot and remote wipe.
Typical use cases
- Computer vision for quality inspection and safety in factories.
- Point-of-sale and inventory systems in retail stores.
- Remote monitoring of energy, utilities and telecom infrastructure.
- Low-latency processing for connected vehicles and logistics.
Common mistakes to avoid
- Treating edge sites as small data centres requiring manual administration.
- Sending all raw data to the cloud, increasing bandwidth costs.
- Underestimating the logistics of replacing failed hardware at remote sites.
Frequently asked questions
How does edge relate to cloud?
Edge complements cloud. Time-sensitive processing happens locally while aggregation, analytics and model training typically run centrally.
Do edge sites need Kubernetes?
Lightweight Kubernetes distributions are popular for edge application management, but simpler options may suit small, single-purpose deployments.
A 90-day action plan
Days 1 to 30: select one use case with measurable value, such as defect detection on a production line, and document latency, bandwidth, resilience and security requirements.
Days 31 to 60: deploy a standard node at a pilot site with zero-touch provisioning, central monitoring and remote update capability.
Days 61 to 90: measure outcomes against the business case, refine the hardware and software standard, and create a rollout plan for further sites.
Questions to ask suppliers
- Is the hardware rated for the temperature, dust and vibration at our sites?
- How are devices provisioned, updated and recovered remotely?
- What happens to applications when connectivity is lost?
- How are devices protected against tampering and theft?
- What are the support and replacement service levels for remote locations?
Key terms explained
- Zero-touch provisioning: automatic configuration of a device when it is first connected.
- Fleet management: centrally managing many distributed devices and applications.
- Inference: running a trained AI model to make predictions on new information.
- Store and forward: holding information locally and sending it when a connection is available.
The bottom line
Processing information close to where it is created enables faster decisions, lower bandwidth costs and continued operation when connections fail. The challenge is managing many distributed sites without local IT staff. Standardised stacks, zero-touch provisioning, central fleet management, designs that tolerate disconnection and strong physical security make large deployments practical. Start with a use case that has clear value, prove it at one site and scale with a repeatable blueprint.
Further reading on edge computing
For authoritative, vendor-neutral guidance on edge computing, see ETSI. You can also browse our free whitepapers.
