How CDNs work and when you actually need one

Content Delivery Networks

A CDN caches your assets at edge nodes globally. A user in Singapore hitting a CDN edge node in Singapore gets the file in ~5ms instead of ~200ms from an origin server in the US. For static assets, this is purely a physics win.

CDN caching is controlled by the same Cache-Control headers your origin sends. Add s-maxage to set a separate TTL for CDN caches:

# Public caches (CDN) cache for 1 year; browsers cache for 1 day
Cache-Control: public, max-age=86400, s-maxage=31536000

# HTML: never cache at CDN (need fresh content), revalidate in browser
Cache-Control: no-cache, s-maxage=0

Modern CDNs (Cloudflare, Fastly, Vercel Edge) also run edge functions — serverless functions that execute at the edge node, reducing API TTFB to under 50ms globally.

When do you need a CDN? Nearly always for static assets. For HTML, a CDN makes sense when your content is cacheable (blogs, marketing sites) but adds complexity for dynamic applications (logged-in user state, personalization).

CDNs also provide DDoS protection, automatic TLS certificates, and HTTP/3 support without server configuration. Cloudflare's free tier covers most static sites entirely.