Making database reads faster: Caching versus using read replicas.

hey folks ive been thinkin a lot about scaling read-heavy apps lately. our main db is startin to feel the strain during peak hours, and im lookin at ways to make read performance better. my first thoughts are caching (like redis or memcached) or settin up read replicas. both have good and bad points, obvs. caching can cut load a lot but adds complexity with invalidation. read replicas spread load well but might have a small replication lag. for those whove dealt with this, whats your experience been? did you find one way way better for some cases? or is a mix usually the way to go? wanna hear your thoughts and any traps to avoid