Don't Panic: Redis Fundamentals

Redis is a server that keeps typed values behind keys, which sounds like a filing cabinet until you notice that the drawers can count, rank, deduplicate, and record events. That is the useful part. A string holds a value or counter. A hash holds named fields. A sorted set holds members with scores. A stream holds ordered entries. The key does not tell Redis what the value means. The data type and command do, so choosing them is the first design decision and not a decorative label applied after the code works.

The usual temptation is to call Redis a cache and walk away briskly before anyone asks what happens to the cached thing. A time to live is a timer attached to a key. When it ends, the key is gone. Memory pressure can also remove keys according to the selected eviction policy. This is excellent when a missing value can be rebuilt from a durable source. It is less excellent when the missing value was secretly the only copy of something important. Computers are very good at enforcing assumptions you forgot you made.

Persistence provides another set of choices. An RDB snapshot records the dataset at a point in time. An append-only file records write operations for replay. Neither option replaces a restore test, because a persistence file that cannot meet the recovery objective is mostly a comforting object with a filename. Replication can improve availability and read capacity, but it is asynchronous by default, so an acknowledged write can still disappear in a failover. This is not Redis being difficult. It is Redis declining to promise a physics-free network.

Redis Cluster adds a further turn of the map. It partitions keys across nodes by hash slot, which helps when one instance no longer fits the measured workload. It also makes key placement and client behavior part of the design. Start with a clear workload instead of treating a cluster as a ceremonial promotion for a database that has reached adulthood.

Read the intro for the full architecture and failure model. Use the slides to see which data structure matches each workload. Keep the cheatsheet nearby when the terms start arriving in packs. The practice reference and exercise make expiry, counters, streams, and inspection commands observable. Field Notes covers the operational costs that tend to appear after the pleasant diagram has left the room. The Reference tab is the route into the exact commands and configuration details.

Don't Panic: Redis Fundamentals

Relevant careers