July 10, 2026

Caching Type

#system-design #cache

Caching can happen at several layers. Using multiple layers reduces latency and prevents every request from reaching the origin database or service.

Client-side caching

Stores data on the user's device, so repeated access is fast and network usage is reduced. Best for static assets and limited offline use.

Examples:

  • Browser cache: stores HTML, CSS, JavaScript, images, and fonts.
  • Application cache: stores API responses, user profile data, and preferences.

Risk: the server cannot immediately remove every client copy, so versioned asset names and appropriate expiration times are important.

Application caching

The server application caches objects that are expensive to retrieve or compute.

  • Local in-memory cache: very fast, but not shared between instances and lost when the process restarts.
  • Distributed cache: stored in systems such as Redis or Memcached; shared by multiple application instances.
    • Suitable for sessions, API results, database records, and computed values.

Risk: cache stampede

Web server and reverse-proxy caching

A web server or reverse proxy, such as Nginx or Varnish, caches complete HTTP responses near the application server. Useful for pages or API responses requested by many users.

  • Can store HTML, API responses, and RSS feeds.
  • Reduces work on the origin server by serving repeated requests directly from cache.

Database caching

This can occur inside the DBMS/database engine, storing frequently used database data or results in memory so the database can answer faster.

  • Buffer/page cache: keeps disk pages in RAM.
  • Query/result cache: saves repeated query outputs.
  • Materialized views: store precomputed query results.
  • Read replicas: reduce load by serving reads, though they are not true caches.
  • Execution plan cache: reuses query plans for repeated statements in some databases.

DNS caching

Stores domain name lookup results so the same hostname does not need to be resolved repeatedly.

  • Can happen in the browser, operating system, local router, recursive resolver, or CDN edge resolver.
  • Uses a TTL to decide how long a DNS record should be kept.
  • If a DNS record changes, cached results may stay visible until the TTL expires.
  • Helps reduce latency before the connection starts, because the client gets the server IP faster.

CDN caching

A CDN stores static assets and cacheable responses at geographically distributed edge locations.

  • Reduces latency by serving content from a location close to the user.
  • Reduces bandwidth and request load on the origin server.
  • Commonly caches images, videos, scripts, stylesheets, downloads, and sometimes public API responses.
  • Content is refreshed through expiration, revalidation, versioned URLs, or explicit cache purging.