CDN & Edge — Java Interview Guide | Cracked Java
Mid

CDN & Edge

What a CDN does, edge cache invalidation, edge compute, anycast vs DNS geo-routing, HTTP/2-3, and TLS termination at the edge.

Prereqs: caching-strategies

A CDN (content delivery network) is a globally distributed fleet of cache servers — points of presence (PoPs) — placed close to users so that content is served from a nearby city instead of your origin on another continent. It is the cache layer that lives outside your data center, and for any system serving a geographically spread audience (media, static assets, even cacheable API responses), it is one of the highest-impact and easiest wins in a design.

Why this matters

Latency is dominated by the speed of light and round-trips. A user in São Paulo fetching from a Virginia origin pays ~120 ms each way; the same object from a São Paulo PoP is ~5 ms. A CDN also offloads the origin — if 95% of requests are served at the edge, your origin sees a fraction of the traffic, which cuts cost and shields you from spikes (a built-in absorber for traffic surges and a first line against DDoS). In an interview, "put a CDN in front of static and cacheable content" is expected; the depth comes from invalidation, edge compute, and routing.

The mental model

  • Edge caching. PoPs cache responses keyed by URL + headers, honoring Cache-Control, ETag, and TTLs. The hard part, as always, is invalidation — TTLs, explicit purge APIs, and surrogate keys that purge groups of related objects at once.
  • Anycast vs DNS geo-routing. Two ways to send a user to the nearest PoP: anycast (the same IP is announced from every PoP; BGP routes the packet to the topologically nearest one) and DNS-based geo-routing (the DNS resolver hands back the IP of a nearby PoP). They differ in failover speed and granularity.
  • Edge compute. Beyond static caching, modern CDNs run code at the PoP — Cloudflare Workers, Lambda@Edge, Fastly Compute — for request routing, auth, A/B tests, personalization, and assembling responses without a trip to origin.
  • TLS termination at the edge. The PoP terminates the user's TLS connection, so the expensive handshake happens close to the user; the edge then reuses warm, pooled connections back to origin.
  • HTTP/2 and HTTP/3. Edges speak modern protocols — HTTP/2 multiplexing over one TCP connection, and HTTP/3 over QUIC (UDP), which removes head-of-line blocking and resumes faster on flaky mobile networks — even if your origin still speaks HTTP/1.1.

What a CDN serves

  • Static assets (JS, CSS, images, video segments) — the classic case, with long TTLs and content-hashed URLs.
  • Cacheable dynamic content — API GETs, HTML fragments — with short TTLs or surrogate-key invalidation.
  • Streaming media — HLS/DASH segments delivered from the nearest edge are what make global video viable.

What the questions cover

The questions explain what a CDN actually does and when it is essential, how cache invalidation works at the edge (TTL vs purge vs surrogate keys), and the routing-and-termination layer — edge compute, anycast vs DNS geo-routing, and why TLS terminates at the edge. Together they take "I'd add a CDN" from a checkbox to a defensible design decision.

Questions

3 in this topic