Redis Best Practices: Naming, Size Limits, Expiration, and Command Pitfalls

Key Naming Design: Readability, Manageability, Brevity

It is recommended to use a colon (:) to separate parts of a key because many Redis clients group keys by colon. For example: apps:app:1, apps:app:2, apps:app:3. Tools like Redis Desktop Manager automatically categorize them under an apps folder.

Value Design: Avoid Big Keys

String values should be kept within 10 KB. Larger values cause a performance breakpoint, as shown in Redis benchmarks. For instance, a 10 KB value with a QPS of 100 000 would consume the full bandwidth of a 1 Gbps network (10 KB × 100 000 = 1 000 MB).

Control Key Lifecycle: Set Expiration

Set an expiration time for every key whenever possible. Without expirations, stale data can accumulate over months, leading to hundreds of gigabytes of unused cache that may eventually cause performance degradation. Expirations enable automatic eviction of obsolete data.

Beware O(n) Commands

Commands like LINDEX and LREM on lists have O(n) complexity, meaning their execution time grows with the number of elements. Since Redis is single‑threaded, a slow command can delay subsequent commands, severely impacting performance.

Disable Dangerous Commands: KEYS, FLUSHDB, FLUSHALL

These commands should be disabled in the configuration (using rename‑command). FLUSHDB and FLUSHALL wipe data entirely. The KEYS command can cause massive performance loss, especially under high concurrency, as illustrated by real‑world incidents.

Prefer Batch Operations for Efficiency

Batch commands fall into two categories:

Native batch commands such as MGET, MSET, HMGET, HMSET, LPUSH with multiple values.

Non‑native batch commands like Pipeline.

Using these commands can dramatically improve performance, especially on single‑node Redis or Sentinel setups. In cluster mode, batch operations are limited by slot boundaries. Keep batch sizes reasonable (generally ≤ 1000) depending on data size.

monitor Command: Limit Usage Time

The monitor command streams all commands executed by the server in real time. Prolonged use can fill the output buffer and increase memory consumption, especially under high concurrency.

afeiMacBook-Pro:redis-3.2.11 afei$ src/redis-cli monitor
OK
1573915193.053188 [0 127.0.0.1:55357] "COMMAND"
1573915197.087383 [0 127.0.0.1:55357] "set" "name" "afei"
1573915217.938838 [0 127.0.0.1:55357] "set" "公众号" "阿飞的博客"

Conclusion

Every technology has trade‑offs; there is no silver bullet. Understanding Redis’s strengths and weaknesses allows you to design a reasonable architecture, write optimal code, and avoid costly mistakes—such as using KEYS in production and incurring financial loss.