Caching

Choose Sentinel when the data fits on one node and you need automatic failover; choose Cluster when the data or request rate exceeds one node and you need horizontal sharding.

Redis is fast because data is in memory, a single event loop over epoll serves thousands of clients with zero lock contention, and Redis 6.0 kept the lock-free model while adding I/O threads for network work only.

A CDN caches static content at the edge close to users, Redis caches dynamic data close to the application, and a semantic cache matches LLM requests by embedding similarity so paraphrased questions reuse a cached answer.

Cache penetration is repeated requests for keys that do not exist always missing the cache and hammering the database, defended by negative caching with a short TTL and a Bloom filter whose 'not present' answer is always correct.

Match the eviction policy to the access pattern, because a cache with a low hit rate costs you a network hop on every miss and still hits the database.

Use SET NX PX (setIfAbsent with TTL) so only one instance acquires the lock, the TTL prevents deadlock on crash, and a token-checked Lua release ensures only the owner unlocks.