CDN in System Design: Edge Servers, PoPs, Anycast & Origin Offload (Visualized)

A Content Delivery Network (CDN) is a geographically distributed network of servers that cache content close to users, so requests are served from a nearby edge location instead of a single distant origin. The result is lower latency, less load on your origin servers, and better availability under traffic spikes.

Every time you load a website's images, scripts, or videos quickly from anywhere in the world, a CDN is almost certainly involved. Providers like Cloudflare, Akamai, and Fastly operate thousands of edge locations worldwide.

Data travels at a finite speed. A round trip from Sydney to a server in Virginia is ~200ms before any processing — and a single page can make dozens of requests. By placing a copy of your content in a Point of Presence (PoP) near the user, a CDN turns that 200ms into a few milliseconds.

A CDN deploys PoPs — clusters of edge servers — in data centers across the globe. When a user makes a request, the CDN routes them to the nearest healthy PoP, usually via anycast: many PoPs advertise the same IP address, and the network delivers the request to the closest one automatically.

Edge servers cache content according to Cache-Control headers and TTLs. Because most requests for popular assets are served from cache, only a small fraction — cache misses and uncacheable requests — ever reach your origin. This origin offload is what lets a modest origin survive massive traffic.

CDNs excel at static assets (images, CSS, JS, video) that can be cached for long periods. For dynamic or personalized content, CDNs still help via dynamic acceleration: terminating TLS at the edge, reusing warm origin connections, optimizing routing, and caching at the edge for short windows or with stale-while-revalidate.

Origin shield: a designated mid-tier cache that all PoPs consult before the origin, collapsing many misses into one origin fetch. Cache key: what identifies a cached object (URL + selected headers). Purge/invalidation: explicitly clearing cached objects when content changes. Hit ratio: the share of requests served from cache — the single most important CDN health metric.

A cache is the general technique of storing copies of data for fast reuse. A CDN is a globally distributed system of caches (edge servers) placed close to users. So a CDN is a cache — specifically a geographically distributed edge cache plus the routing to reach it.

Yes, even when content can't be cached. Through dynamic acceleration — TLS termination at the edge, optimized backbone routing, and reused origin connections — a CDN reduces round-trip overhead. It can also apply short TTLs or stale-while-revalidate to safely cache semi-dynamic responses.

A load balancer distributes traffic across servers, usually within one region/data center. A CDN distributes cached content across many global locations to serve users from nearby. They're complementary: a CDN sits at the edge in front of your origin, which may itself sit behind a load balancer.