You add an in-process cache on each app server in front of a shared Redis. What do you gain and what do you pay?

Answer

Gain: hits with no network hop at all, and less load on Redis. Pay: each server holds its own copy, so memory is duplicated and invalidation must reach every server (or you accept staleness up to a short TTL). Keep the local layer small and its TTL short.

Review this in your daily deck → All cards in Caching