Edge Computing Explained: Processing Near the Data
Edge computing puts computation near the place where data is created, so a decision can be made in milliseconds instead of travelling to a distant data centre.
In this guide
Cloud computing centralised processing in large data centres. Edge computing pushes some of that work back out towards cameras, vehicles, factory floors, retail stores and cell towers.
It is not a replacement for the cloud. It is a complement: local processing for what must be immediate, cloud processing for everything that benefits from scale, history and coordination.
Why move compute to the edge
Physics sets a floor on latency. A round trip to a distant region costs tens to hundreds of milliseconds, which is fine for loading a page and unacceptable for a braking decision or an industrial safety interlock.
Bandwidth and cost matter too. Sending every frame from a hundred cameras to the cloud is expensive and often pointless when only a few seconds contain an event. Filtering locally sends the signal and drops the noise.
- Latency: react within milliseconds without a round trip.
- Bandwidth: upload summaries instead of raw streams.
- Resilience: keep working when the backhaul fails.
- Privacy: keep sensitive footage or readings on site.
- Cost: reduce cloud egress and storage charges.
What an edge deployment looks like
Edge sits on a spectrum. Device edge runs directly on sensors and cameras. Near edge uses a local server or gateway in a store, factory or cell site. Regional edge places small data centres closer to users than a central region.
Workloads are usually containerised so the same application can run in several locations. Management, updates and observability are centralised, because hundreds of remote sites cannot be visited by hand.
The trade-offs nobody mentions in the keynote
Distributed systems are harder than centralised ones. Every edge site needs power, cooling, physical security and a plan for when it fails. Software must tolerate intermittent connectivity and partial data.
Edge hardware is heterogeneous and often unpatched. Model and application updates must be signed and roll back safely, and someone has to own the fleet after the pilot ends.
Where edge computing fits best
Computer vision on production lines, autonomous vehicles, retail analytics, medical devices, content delivery and 5G-enabled industrial control all justify local processing. Each combines a hard latency or bandwidth constraint with a clear cost of failure.
A useful test is to measure the real latency requirement first. If the answer is hundreds of milliseconds, a well-designed cloud service will usually do the job.
Key facts at a glance
Frequently asked questions
Is edge computing the same as a content delivery network?
A CDN is one form of edge computing focused on caching content close to users. Edge computing is broader and includes running application logic and analytics locally.
Does edge computing replace the cloud?
No. Edge handles local, time-sensitive work while the cloud provides central storage, coordination, training and long-term analytics.
What is edge AI?
Running machine learning models on or near the device rather than in a data centre, usually with smaller optimised models so inference can happen locally.