Field notes on Redis — the single-threaded event loop, the two persistence models, the data structures that turn a 10× memory win into one line of code, and the cache-stampede patterns every team needs before the first outage.
7 articles · updated regularly
A single popular cache key expired at peak traffic and three thousand requests all missed at once, all ran the same expensive query, and the database fell over. The fix isn't a longer TTL. It's making sure exactly one request rebuilds the value.
How a single product page took down our Redis cluster on the busiest night of the year — and the one-line fix that came too late.
Cluster scales horizontally, but multi-key commands come with new rules. Hash tags are the escape hatch.
Sorted sets, streams, hyperloglog, bitmaps — picking the right one is usually a 10x win on memory or latency.
Everyone has been paged by a cache miss storm. Here are the four battle-tested patterns for preventing one.
Snapshots vs append-only: what each one guarantees, what each one loses, and how to combine them safely.
Why Redis commits to a single thread, how the I/O model really works, and where multithreading slips in.