Chapter 35: Caching Patterns — Read-Through, Write-Behind
Caching reduces database load by serving frequently-accessed data from fast storage (Redis, in-memory). But cache invalidation is one of the hardest problems in computer science.
Cache-Aside (Lazy Loading)
1. App checks cache: GET user:123
2. Cache MISS → query database
3. Store result in cache: SET user:123 (with TTL)
4. Return to client
Next request:
1. App checks cache: GET user:123
2. Cache HIT → return immediately (no DB query)
// Pseudocode: cache-aside pattern
user = cache.get("user:123")
if user == nil {
user = db.query("SELECT * FROM users WHERE id=123")
cache.set("user:123", user, ttl=300) // 5 min TTL
}
return userWrite-Through
Write to cache AND database simultaneously. Cache is always up-to-date.
// Write-through: update both
db.update("UPDATE users SET name='Bob' WHERE id=123")
cache.set("user:123", updated_user) // cache always freshWrite-Behind (Write-Back)
Write to cache first, asynchronously flush to database later. Fast writes but risk of data loss if cache crashes.
Cache Invalidation Strategies
| Strategy | How | Trade-off |
|---|---|---|
| TTL (Time-To-Live) | Cache expires after N seconds | Simple but stale data for up to TTL |
| Event-based | Invalidate on write (pub/sub) | Fresh but complex |
| Version-based | Include version in cache key | Never stale but more cache misses |
Common Pitfalls
⚠️ Cache Problems
- Cache stampede: TTL expires, 1000 requests hit DB simultaneously. Fix: lock + single refresh.
- Stale data: cache shows old value after update. Fix: invalidate on write.
- Cold start: empty cache after restart, DB gets hammered. Fix: warm cache on deploy.
- Over-caching: caching everything wastes memory. Cache only hot, expensive queries.
Key Takeaways
- Cache-aside: simplest pattern, app manages cache explicitly
- Write-through: cache always fresh, slightly slower writes
- TTL is the simplest invalidation — accept some staleness
- Only cache data that's read frequently and expensive to compute
- "There are only two hard things: cache invalidation and naming things"