Edge versus fog versus cloud
In short: Cloud computing centralises large, elastic resources; edge computing acts close to devices; fog computing describes the layer in between that aggregates and coordinates several edge sites. Real systems use all three, placing each task where latency, bandwidth, resilience and cost requirements are best met.
Three terms, one continuum
It is easiest to see cloud, fog and edge as points along a line from the device to a large data centre.
- Cloud: large shared data centres offering elastic compute, storage and managed services. Best for heavy analytics, long-term storage, training AI models and anything that is not time-critical.
- Edge: processing on or very near the device, such as in a camera, gateway or on-site server. Best for instant reactions and filtering data.
- Fog: an intermediate layer, for example a site or district server that collects data from many edge nodes, coordinates them and passes summaries to the cloud. The term was popularised in the IoT world, and an industry reference architecture for fog was later adopted as an IEEE standard (IEEE 1934).
In practice many vendors now use "edge" for everything outside the central cloud, so "fog" appears less often. You will still see it in IoT courses and exams, as in IoT edge, fog and cloud.
Comparing the tiers
| Factor | Edge | Fog | Cloud |
|---|---|---|---|
| Distance from devices | On or next to them | Local site or area | Regional or global data centre |
| Response time | Fastest | Fast | Slowest of the three |
| Compute and storage | Small | Medium | Very large, elastic |
| Works without WAN | Yes | Usually | No |
| Management effort | High (many nodes) | Medium | Lowest per workload |
Deciding where a workload runs
- Latency: how fast must the system respond? Milliseconds point to edge; seconds or minutes allow cloud.
- Data volume: is it cheaper to process locally and send summaries?
- Resilience: must the function keep working during a WAN outage?
- Data rules: must data stay on site or in a country?
- Compute needs: does the job need more power than an edge node has, such as model training?
- Operations: can you patch, monitor and physically protect the location?
Worked example: a smart building
Temperature sensors and door controllers sit on each floor. An edge controller on each floor reacts instantly, for example unlocking doors during a fire alarm even if the network is down. A fog server in the building's plant room combines data from all floors, balances heating and keeps a week of logs. The cloud platform stores years of history, compares many buildings and trains a model to predict energy use, then sends updated settings back down.
Tip: security policy must follow data across all tiers. Encrypt links between tiers, authenticate every node and decide which tier is trusted to issue commands to devices.
Practice
Where should AI model training for a fleet of buildings usually run?
What does the fog tier typically do?
Name one reason to keep a door control decision at the edge.
Key takeaways
- Cloud, fog and edge are points on a line from data centre to device.
- Edge acts fast and locally; cloud offers scale; fog coordinates in between.
- Choose placement by latency, data volume, resilience, data rules and compute needs.
- Most real designs combine all three tiers.
- Security must cover every tier and the links between them.
Original content written by Mohamed Fayas for FM TecHub. © 2026 FM TecHub. You're welcome to share a link to this page; please don't copy the text elsewhere without permission.