Edge vs Cloud
| Model | Where Work Runs | Best For |
|---|---|---|
| Central cloud | Large regional data centers | Databases, analytics, coordination, durable storage |
| Edge computing | Near users, devices, or access networks | Low-latency responses, caching, filtering, local decisions |
| Device edge | On phones, cameras, routers, machines, sensors | Immediate local control and offline behavior |
The Latency Motivation: Speed of Light
Light travels through fiber optic cable at roughly 200,000 km per second — about two thirds the speed of light in a vacuum. A round trip from New York to London and back covers about 11,000 km of fiber distance, which translates to a minimum of around 55 ms of propagation delay before accounting for routers, queuing, or processing. That physical limit cannot be engineered away. The only way to reduce it is to move computation or cached data geographically closer to the user. This is the fundamental motivation behind edge computing: not every workload needs to run in a central data center 3,000 miles away when it could run 30 miles away instead.
Categories of Edge
Edge computing is not a single architecture. Several distinct categories have developed, each with different scale, ownership, and use cases.
- CDN edge — Servers at internet exchange points and ISP facilities that cache static content (images, video, JavaScript, fonts) close to users. The origin server only delivers content once; the CDN edge serves subsequent requests locally.
- CDN compute edge / serverless at the edge — Platforms like Cloudflare Workers and Fastly Compute run small JavaScript or WebAssembly programs at CDN edge nodes. This moves request logic (personalization, A/B testing, authentication, response shaping) out of the origin entirely.
- MEC (Multi-access Edge Computing) / 5G edge — Compute infrastructure co-located with mobile base stations or at carrier facilities. 5G networks enable ultra-low latency (sub-10 ms) paths to edge compute for use cases that require near-real-time processing of mobile device data.
- IoT edge — Processing at or near the device: inside a factory, warehouse, store, or building. A machine vision system that inspects products on an assembly line cannot afford 100 ms of round-trip latency to a cloud; it processes frames locally and sends only results or anomalies upstream.
- On-device edge — The device itself performs inference or decision-making. Modern smartphones, cameras, and embedded controllers contain neural processing units (NPUs) specifically for running ML models locally.
Serverless at the Edge
Cloudflare Workers, Fastly Compute, and similar platforms allow developers to deploy small functions that run inside CDN edge nodes distributed across dozens or hundreds of cities. A request from a user in Tokyo hits a nearby edge node rather than an origin server in Virginia. The function can read and write to distributed key-value stores, fetch from APIs, modify HTTP responses, and enforce security policies — all without a full VM or container startup time. Latency for simple operations can be under 5 ms from most populated areas. The trade-off is a constrained execution environment: no long-running processes, limited CPU time per request, and restricted access to OS-level resources.
Real-World Applications
- Video processing — Live streaming platforms transcode video at edge nodes near broadcast sources to reduce upload latency and avoid backhauling raw high-bitrate streams to a central region.
- Autonomous vehicles — Self-driving systems process sensor data on-vehicle in real time. Some decisions (emergency braking) happen in milliseconds on the vehicle's own compute. Map updates and fleet-level learning flow to and from cloud systems asynchronously.
- Smart factories — Industrial IoT systems use local edge servers to run quality control, anomaly detection, and process control algorithms. A factory floor cannot tolerate WAN outages stopping production lines.
- Retail and logistics — Edge compute in stores and warehouses runs inventory vision systems, checkout automation, and route optimization without cloud dependency for each transaction.
- Gaming — Cloud gaming and multiplayer matchmaking use edge locations near players to minimize input lag. A 10 ms reduction in server RTT is meaningful for competitive gaming.
Trade-offs vs Centralized Cloud
Edge computing improves latency and reduces backhaul traffic, but it introduces real complexity that centralized systems avoid.
| Concern | Centralized Cloud | Edge Computing |
|---|---|---|
| Data consistency | Easier — single authoritative source | Harder — distributed copies must synchronize |
| Deployment | One or a few regions to update | Hundreds of locations to update consistently |
| Security patching | Centralized update path | Every edge node needs patching, some may be physical hardware in untrusted locations |
| Observability | Centralized logs and metrics | Logs distributed across many locations, harder to correlate |
| Hardware management | Managed by cloud provider | Physical edge hardware requires maintenance and replacement |
How Edge Complements Rather Than Replaces Cloud
Edge and cloud are not competitors — they are partners in a layered architecture. The edge handles what needs to be fast, local, or bandwidth-efficient: caching, filtering, inference, session handling. The cloud handles what needs to be durable, globally consistent, or computationally large: databases, analytics pipelines, model training, billing systems, user account management. A well-designed edge system offloads the cloud by handling repetitive or latency-sensitive work locally, while still relying on the cloud for ground truth. Treating edge as a replacement for cloud typically leads to operational problems around consistency and management complexity.
How It Shows Up in Speed Tests
A speed test often selects a nearby server for the same reason edge systems exist: distance and routing affect results. A nearby server usually reduces latency and avoids long transit paths. A far-away server can reveal how your connection behaves beyond your local access network.
Frequently Asked Questions
What is edge computing in simple terms?
Edge computing means running compute or storing data closer to users, devices, or sensors instead of sending everything to a distant central data center.
Is a CDN edge computing?
A CDN is one common edge pattern for content delivery. Modern edge computing can go beyond caching by running logic, security checks, personalization, or inference near users.
Does edge computing replace cloud computing?
No. Edge and cloud usually work together. The edge handles latency-sensitive or local work, while central cloud systems handle coordination, storage, analytics, and large-scale control.