For senior engineers, a Content Delivery Network (CDN) is more than just a black box that makes sites faster. It is a distributed system designed to trade storage for latency, bringing data physically closer to the user to bypass the limitations of the speed of light.
1. Core Architecture: Origin, PoP, and Edge
The architecture consists of three distinct layers:
- Origin Server: The 'Source of Truth'. This is your S3 bucket, EC2 instance, or on-premise rack housing the master copy of the application assets.
- Point of Presence (PoP): Physical data centers located in strategic geographic hubs (internet exchange points).
- Edge Servers: The actual machines within a PoP that handle HTTP termination, SSL/TLS handshakes, and caching logic.
2. Caching Strategies: Push vs. Pull
Deciding how data gets to the edge is a critical design decision.
- Pull Zone (Standard): The edge server checks its cache when a request arrives. If it's a MISS, it fetches from the origin, serves the user, and caches the result. Best for high-traffic, frequently updated sites.
- Push Zone: The engineering team explicitly uploads content to the CDN storage before user requests occur. Best for large, immutable files like software patches or installers (500MB+).
3. Anycast Routing
How does a user in London connect to the London PoP while a user in New York connects to the NY PoP using the same IP address? This is achieved via IP Anycast and BGP (Border Gateway Protocol).
Multiple geographically dispersed servers advertise the exact same IP address. Routers on the open internet choose the shortest topological path (lowest hop count) via BGP.