# Caching Strategies

> Cache effectively — in-memory, distributed, HTTP caching — with proper invalidation.

- Skill: `rahulrachhoya/caching-strategies` (Agent Skill)
- Install (CLI): `npx skillmds@latest add rahulrachhoya/caching-strategies`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rahulrachhoya/caching-strategies/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: RahulRachhoya (https://skillmd.com/u/rahulrachhoya)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/rahulrachhoya/caching-strategies

---


# Caching Strategies

## Cache Levels
| Level | Storage | Speed | Use Case |
|-------|---------|-------|----------|
| L1 (in-memory) | Application RAM | <1ms | Hot data, repeated queries |
| L2 (distributed) | Redis/Memcached | 1-5ms | Shared across instances |
| L3 (HTTP/CDN) | Browser/CDN | 10-100ms | Static assets, API responses |

## Caching Patterns
- **Cache-aside**: read from cache, miss → read DB → write to cache → return
- **Write-through**: write to cache first, then DB
- **Write-behind**: write to cache, async write to DB
- **Cache invalidation**: the hardest problem — use TTLs, version keys, or event-driven purges

## HTTP Caching
- Use `Cache-Control` headers: `public`, `private`, `max-age`, `no-cache`, `no-store`
- Use ETags for conditional requests
- Set appropriate TTLs: static assets (1 year with hash in URL), API responses (seconds to minutes)

## Rules
- Cache the result of expensive operations (DB queries, API calls, computation)
- Never cache sensitive data (PII, auth tokens)
- Always set a TTL (even if long) — prevent stale data
- Monitor cache hit rates — low hit rate means wrong cache strategy
- Have a cache warming strategy for critical data