One cache key gets 1 million reads a second, more than a single cache node can serve. What are the usual fixes?

Answer

Copy the key: store it as key#1 … key#N on different nodes and read a random copy (writes update all of them). Or add a small in-process cache on each app server with a TTL of a second or so, so most reads never leave the server. Both trade some staleness for spreading the load.

Review this in your daily deck → All cards in Caching