2 / 5 Cache Patterns & Write Policies

Every Sixty Seconds

One extremely hot key (the homepage feed) is cached with a 60 second TTL. Latency graphs show a clean p99 spike almost exactly once a minute, when the entry expires and the next requests fall through to a 400 ms database query. Which change targets this directly?

A

Add read replicas to the cache cluster so the hot key is served from more nodes at once

B

Refresh-ahead: recompute the entry in the background just before its TTL elapses

C

Raise the TTL to one hour so the expensive query runs sixty times less often per key

D

Switch the feed to a write-through cache so every write keeps the cached copy current