CDN

A CDN is a network of servers placed in data centers around the world, often called edge locations or points of presence (PoPs). Each edge server keeps cached copies of content that originally lives on your own server, which the CDN calls the origin.

When someone in Tokyo loads your website, they do not pull every image and script from your server in Virginia. The CDN routes them to a nearby edge server, maybe in Tokyo or Osaka, that already holds those files. The data travels a few hundred kilometers instead of crossing the Pacific, so the page loads faster.

Large CDNs are huge. Cloudflare advertises a presence in over 300 cities. Akamai runs hundreds of thousands of servers across thousands of locations. The point of all that hardware is simple: keep a copy of your content physically close to as many people as possible.

Two pieces do the heavy lifting: request routing and caching. When a user requests a file, the CDN uses DNS or anycast routing to send them to the closest healthy edge server, usually measured by network distance rather than raw geography.

On the first request for a file, the edge server does not have it yet. This is a cache miss. The edge fetches the file from your origin, sends it to the user, and stores a copy. Every later request for that file from that region is a cache hit, served directly from the edge with no trip back to origin.

How long a file stays cached is controlled by HTTP headers, mainly Cache-Control and the file's TTL (time to live). A logo with a one year TTL almost never hits your origin again. When you ship a new version, you either change the filename or send a purge command that tells edge servers to drop the old copy.

CDNs were built for static content like images, CSS, JavaScript, fonts, and video segments. Modern ones also do more: terminating TLS, compressing responses, blocking DDoS attacks, and running code at the edge with platforms like Cloudflare Workers or AWS Lambda@Edge.

Reach for a CDN any time you serve the same files to many users in different places: a marketing site, a video platform, an app's static assets, or software downloads. The bigger your geographic spread, the bigger the win. A site whose users are all in one city sees far less benefit.

The main trade-off is staleness. Aggressive caching means an edge server can keep serving an old file after you have updated the origin. You manage this with sensible TTLs, versioned filenames, and cache purges, but a misconfigured cache can show users yesterday's content or, worse, leak a private response to the wrong person.

CDNs are weaker for content that is unique per user or changes every second, like a personalized dashboard or live stock prices. You can still route those requests through the CDN for TLS and DDoS protection, but you usually mark them as uncacheable so they pass straight to origin.

Other costs to weigh: most CDNs bill per gigabyte of egress and per request, debugging gets harder because a cached bug can persist after a deploy, and you are trusting a third party with your traffic. For the vast majority of public web content, the latency and reliability gains are worth it.

Netflix is the textbook case. Instead of streaming every movie from a central data center, Netflix built its own CDN called Open Connect and places storage appliances, called OCAs, directly inside internet service provider networks and at internet exchange points.

Each appliance is preloaded overnight with the shows most likely to be watched in that region. When you press play, the video streams from an OCA that might be in the same building as your ISP, often just one network hop away. That is why a 4K stream starts in a second or two without buffering.

This design saves enormous bandwidth. Netflix can account for a large share of evening internet traffic in many countries, and Open Connect keeps that traffic off expensive long haul links by serving it locally. It is the same core idea as Cloudflare or Akamai, just owned and tuned end to end by one company for one workload.