Redis — Trivia & Interesting Facts¶
Surprising, historical, and little-known facts about Redis.
Redis stands for "Remote Dictionary Server"¶
Salvatore Sanfilippo (antirez) created Redis in 2009 while working on a real-time web analytics startup called LLOOGG in Italy. He needed a fast way to process web page views and found existing databases too slow. The name "Remote Dictionary Server" reflects its origin as a networked key-value (dictionary) store.
Redis is single-threaded by design and this is considered a feature¶
Redis processes all commands on a single CPU core using an event loop. This eliminates locking overhead and makes individual operations inherently atomic. Despite being single-threaded, Redis can handle over 100,000 operations per second on modest hardware because most operations are O(1) and memory-bound. Multi-threaded I/O was added in Redis 6.0 (2020) but only for network handling, not command execution.
The Redis license change in 2024 fractured the community¶
In March 2024, Redis Ltd. changed Redis's license from BSD-3-Clause to a dual RSALv2/SSPL license, preventing cloud providers from offering Redis as a managed service. The Linux Foundation immediately forked Redis as "Valkey," backed by Amazon, Google, Oracle, and Ericsson. This was one of the most significant open-source license disputes since Elasticsearch's similar move in 2021.
Twitter used Redis to store the timelines of over 300 million users¶
Twitter's timeline service stored recent tweets for every user in Redis, with the entire "fan-out on write" architecture depending on Redis performance. At peak, Twitter's Redis deployment handled millions of operations per second. The company's extensive Redis usage led them to contribute multiple improvements back to the project.
Redis persistence has two modes and both have surprised operators¶
RDB snapshots create point-in-time dumps via fork(), which can cause memory usage to briefly double due to copy-on-write. AOF (Append-Only File) logs every write operation but requires periodic rewriting to prevent the file from growing unboundedly. Many operators have been surprised by RDB's memory spike or AOF's disk growth, leading to production incidents.
Redis Cluster took 6 years from proposal to stable release¶
Redis Cluster, providing automatic sharding across multiple nodes, was first proposed in 2009 and reached stable release in Redis 3.0 (April 2015). The long development time reflected the fundamental difficulty of adding distributed capabilities to a system designed as a single-node, single-threaded server. Even after release, many large deployments preferred client-side sharding with tools like Twemproxy.
The KEYS command is explicitly documented as dangerous in production¶
Redis's KEYS * command scans every key in the database and blocks the entire server during execution. On a Redis instance with millions of keys, this can cause a multi-second freeze that triggers cascading timeouts in dependent services. The Redis documentation warns: "Consider KEYS as a command that should only be used in production environments with extreme care." SCAN was introduced as the safe alternative.
Redis Lua scripting provides server-side atomicity that changed how people use Redis¶
Redis 2.6 (2012) introduced Lua scripting with the EVAL command, allowing complex multi-step operations to execute atomically on the server. This turned Redis from a simple cache into a coordination primitive — enabling distributed locks, rate limiters, leaderboards, and state machines that would otherwise require multiple round trips and race-condition-prone client logic.
Sentinel was built after a single Redis failure could take down entire applications¶
Redis Sentinel, the HA monitoring and failover system, was introduced in Redis 2.8 (2013). Before Sentinel, Redis had no built-in failover — if the primary died, manual intervention was required. Production teams at companies like Instagram and Pinterest had built custom monitoring scripts, and the lack of native HA was Redis's most requested feature for years.