High-performance Postgres + Redis architecture: when does caching actually start paying off vs just adding complexity?

Irene López · 0 likes

cache invalidation is still one of the two hard problems in computer science for a reason, ripped out a redis layer at my last job for the exact same reason you describe

Mia Bauer · 0 likes

disagree a bit for read heavy workloads with expensive aggregation queries redis absolutely pays for itself even at moderate scale. depends entirely on your query shape

Ethan Brooks · 0 likes

yeah this, OP's case sounds like it was premature caching not that caching itself is bad

Imogen Clarke · 0 likes

postgres with proper indexes and materialized views gets you further then people think, agree with removing redis if your qps isnt actually that high

Sebastián Rojas · 0 likes

the real answer is always 'it depends on your access pattern' but nobody wants to hear that in a hot take thread lol

Maksym Bondar · 0 likes

we use redis purely for session storage and rate limiting now, not as a query cache at all. much simpler mental model and way fewer bugs

Thandi Mokoena · 0 likes

pgbouncer alone fixed like 80% of our perceived 'we need a cache' problems, connection overhead was the actual bottleneck not query speed

Valentina Ruiz · 0 likes

at faang scale redis isnt optional its mandatory, but 99% of us arent operating at that scale and are just cargo culting the architecture diagrams we see on twitter

Mehdi Benali · 0 likes

this is the real talk nobody says out loud enough

Emir Kaya · 0 likes

read replicas solved more of our scaling problems than a cache layer ever did, cheaper to reason about too