Both are caches. Both reduce load on your origin. They live at different layers of the stack, cache different things, and solve different problems. A candidate who says "we'll use a cache" and can't distinguish them is punting on an important design choice.
Going from user to database, cache opportunities exist at every layer:
Each layer catches what the layer above missed. A request that hits the browser cache costs nothing. A request that misses everything and hits disk is thousands of times more expensive.
| Aspect | CDN | Application cache (Redis) |
|---|---|---|
| Location | ~300 PoPs globally | Your datacenters |
| Stores | HTTP responses — static assets, public API responses | Arbitrary structured data — sessions, counters, computed results |
| Latency to user | ~10–50ms (near edge) | App server → Redis: 1–3ms, then network to user |
| Cache key | URL + varied headers | Arbitrary keys |
| Invalidation | API purge or versioned URLs | DEL key |
| Who runs it | Cloudflare / AWS / Akamai | You |
| Cost model | Per-request + bandwidth | Per-GB memory per hour |
| Best for | Public, cacheable, geographically distributed reads | Private, compute-expensive, per-user data |
CDN wins for: images, JavaScript bundles, CSS, fonts, public API responses (leaderboards, product catalogs), video segments. Anything where the same bytes get sent to thousands of users. The CDN absorbs the fan-out so your origin doesn't.
App cache wins for: per-user sessions, personalized feeds, rate-limit counters, hot DB rows, precomputed aggregations, leaderboard state (writes fast, reads fast). Anything that's per-user or frequently mutated.
The mistake: trying to put per-user data behind a CDN. Cache hit rate ≈ 0 (each user is unique). You pay CDN cost for no benefit. Put per-user data in app-level cache close to your backend.
They're not alternatives — they stack. A well-designed page uses both:
Your origin sees < 5% of total traffic. Of that, 90%+ is served from Redis. Your database sees < 1%. This is what scaling looks like.
Browser cache catches 20%. CDN catches 60% of what's left. App cache catches 90% of what's left. DB sees ~3%. Disk reads are rare events.
Video segments on Open Connect (their own CDN). EVCache (Netflix's Memcached variant) for session, profile, recommendations.
Storefront HTML + assets through Cloudflare. Cart, session, inventory counts in Redis. Two-layer defense against the DB.
Article HTML cached at Fastly with 2-min TTL. Breaking news flushes the cache via API. Origin sees < 1% of reads.
Every response is per-merchant, per-account — no CDN benefit. Aggressive Redis caching of hot rows. Origin sees every request but serves fast.
URL shortener: CDN caches redirects for top URLs (shared across users). News feed: Redis for per-user feed; CDN for static assets. YouTube/Netflix: CDN absolutely critical for video; Redis for session/personalization.
