Edge Computing
Move the computing itself, not just the data, close to the user.
A regular CDN was just the start
A traditional CDN serves static content from the edge. Images, CSS, JavaScript, video. That covers most of the bytes a web app sends to a user, but not the logic.
What if you could also run actual code at every edge location? Auth checks, personalization, A/B testing, redirects, image resizing. All of it happening at the edge, close to the user, without ever talking to your origin server.
That is what edge computing is. Platforms that do this include Cloudflare Workers, AWS Lambda@Edge, Vercel Edge Functions, and Fastly Compute@Edge. You upload a small piece of code. The platform runs that code at every point of presence in their network. Your code runs near the user, no matter where they are.
What you actually use this for
Auth. Check a token at the edge. If it is invalid, return a 401 error immediately. Your origin never sees unauthenticated traffic.
Personalization. Read a cookie at the edge, swap in a personalized banner, then send the result. The cached page stays generic, but every user sees their own customization.
A/B testing. Read a cookie or hash the user ID, decide which version of the page to serve, and route. No round trip to a separate experiment service.
Image transformation. Upload one large image to your origin. The edge resizes it on the fly per request and caches the result. Origin stores one big version. Edge serves dozens of sizes.
Geographic routing. Send EU users to your EU origin and US users to your US origin. Just a few lines of code at the edge.
Why this is a big deal
Edge computing erases the old line between "static" content and "dynamic" content. You used to have to pick. Cache the page for everyone (fast but generic). Or build the page per user (slow but personal). With edge functions, you can do both. The base page is cached. The customization happens at the edge in milliseconds.
There is a catch though. Edge environments are limited on purpose. Your code runs inside small lightweight runtimes like V8 isolates, WebAssembly, or MicroVMs. Not full containers. You get memory limits. Strict execution-time limits, usually 50 milliseconds or less. No filesystem. Limited language support.
The reason is that your code runs as part of every single request. It has to be fast and predictable. If you need to query a database or do heavy work for several seconds, do it on your origin server. The edge is for quick decisions, not heavy lifting.