Why Cache Invalidation Is One of Software Engineering's Hardest Problems

A developer's 30-day system design study reached Day 3 with an in-depth look at caching, covering where it appears across modern architectures — from browser and CDN layers to in-memory stores like Redis and built-in database caches. The article explains three core cache-writing strategies: cache-aside, write-through, and write-behind, each involving distinct trade-offs between consistency, latency, and durability. Invalidation emerges as the central challenge, since determining whether cached data is still accurate — without constantly querying the source — largely defeats the purpose of caching. Common invalidation approaches include TTL expiry, explicit invalidation on writes, and event-based invalidation via message queues, each introducing its own failure risks. The piece also highlights a concurrency risk known as a cache stampede, where simultaneous cache misses flood the underlying database at once.