Edge Computing
Edge Computing
A model that moves data processing to nodes physically close to the source — a cell tower, a local server, or the device itself — rather than routing everything to a distant cloud data center.
Edge computing is a response to one hard constraint: the speed of light. Data in fiber travels at roughly 200,000 km/s — about two-thirds the speed of light in vacuum. A round trip to a cloud data center 1,000 km away takes at least 10 ms just for the physics, before any queuing or processing delay is added. For applications requiring single-digit or sub-millisecond responses, that is simply too slow. Edge nodes — small computing facilities placed at or near the network edge — process data locally and return results before a round trip to a remote data center would even complete.
CDN edge vs compute edge
The term "edge" covers two related but distinct concepts. CDN edge (content delivery network) involves caching static assets — images, videos, scripts — at distributed points of presence (PoPs) close to end users. When you load a webpage, the images may come from a CDN node 20 km away rather than a origin server 5,000 km away. Compute edge goes further: instead of just serving pre-cached files, it runs dynamic application code at those same PoPs. Cloudflare Workers, Fastly Compute, and AWS Lambda@Edge execute serverless JavaScript or WebAssembly at hundreds of locations worldwide, letting application logic run within milliseconds of the user rather than in a single central region.
Edge computing categories
| Category | Location | Typical latency | Example |
|---|---|---|---|
| Cloud | Central data center | 20–100+ ms | AWS us-east-1, Azure West Europe |
| CDN edge | ISP or carrier PoP | 2–10 ms | Cloudflare, Akamai, Fastly |
| MEC / 5G edge | Co-located with cell tower | 1–5 ms | AWS Wavelength, Azure Edge Zones |
| IoT / on-premises edge | Local gateway or server | <1 ms | Industrial PC, Raspberry Pi cluster |
| On-device | The end device itself | Sub-millisecond | Smartphone NPU, autonomous vehicle ECU |
Real-world use cases
Edge computing is not theoretical — it is already powering production systems:
- Video transcoding: live stream encoding happens at edge nodes close to the broadcaster, reducing upload bandwidth and latency before distribution
- Autonomous vehicles: obstacle detection and braking decisions happen on-vehicle or at roadside MEC nodes — a cloud round trip at 50 ms is ten times longer than a typical human reaction threshold for emergency braking
- Retail analytics: in-store cameras process foot traffic and shelf analytics locally, sending only summary data to the cloud instead of raw video streams
- Industrial IoT: factory PLCs and robots need sub-10 ms control loops that cloud connectivity cannot guarantee
- Serverless edge workers: Cloudflare Workers handles A/B testing, authentication, and personalisation at the CDN layer before requests reach the origin
Trade-offs vs central cloud
Edge computing introduces real operational complexity. Consistency is harder — state must be synchronised across many edge nodes rather than managed in one place. Security patching must be applied to hundreds or thousands of distributed nodes rather than a handful of data centre racks. Management overhead increases: monitoring, logging, and deployment pipelines must handle a geographically dispersed fleet. Edge hardware also has limited compute capacity compared to a cloud region — workloads that need terabytes of RAM or GPU clusters still belong in the cloud.
Edge complements, not replaces, the cloud
Most edge deployments use a hybrid model: latency-sensitive processing happens at the edge, while long-term storage, batch analytics, model training, and compliance-sensitive workloads stay in the central cloud. A real-time fraud detection model might run inference at the edge (decision in <5 ms) while retraining on aggregated data happens nightly in a cloud data centre. The edge and cloud are complementary layers, not competitors.
Frequently Asked Questions
What is the difference between edge computing and cloud computing?
Cloud computing centralizes processing in large data centers that may be thousands of kilometres away. Edge computing moves processing to nodes close to the data source. The tradeoff is lower latency vs. less centralized compute power.
How does edge computing reduce latency?
Physics limits data to ~200,000 km/s in fiber. A cloud data center 1,000 km away adds at least 5 ms each way. An edge node 10 km away adds under 0.1 ms. For sub-millisecond response requirements, edge is the only viable option.
How is edge computing related to CDNs?
CDNs are an early form of edge computing — caching static content at distributed nodes. Modern edge computing runs dynamic code and AI inference at those same nodes, not just cached files. Cloudflare Workers and AWS Lambda@Edge are examples.