Caching Strategies: Cache Aside, Write Through, and Cache Invalidation

We walk the request path for the four cache strategies — cache aside, read through, write through, write back — tracing which component owns the cache and who gets blocked when it misses. The interesting part is invalidation: we simulate the race where the writer updates the database, the reader fills the cache with the old value, and the two orders of operations that prevent it. We also measure what each strategy does to p99 latency and to write durability, so the choice between them is visible as cost, not opinion.

Topics covered:

• Cache aside: application-driven loads and the stampede problem
• Read through and write through, and where the lock lives
• Write back and the durability window
• The stale-read race, and TTLs as a safety net
• Cache eviction and why it can behave like a crash