The rise of Redis and the 2026 in-memory data store competition

The engineering origin and commercial growth

Salvatore Sanfilippo began writing Redis in 2008 to address performance bottlenecks in his LLOOGG real-time web analytics service. The system needed to push analytics events and retrieve page views when volume increased. His 2009 version 0.1 included basic commands like set, get, and del. A version 0.2 released two weeks later added lists and fork() based persistence. In 2010, VMware hired Sanfilippo to work full-time on the project. In 2011, Ofer Bengal and Yiftach Shoolman founded Garantia Data, which later became Redis Labs. This company acquired MyRedis in 2013 and launched Redis Enterprise Cloud in February 2013 to serve enterprise clients. Sanfilippo joined Redis Labs in 2015 as the lead developer and official sponsor for the open-source project. The company expanded through the 2016 launch of a module marketplace and the 2019 acquisition of RDBTools. This company grew through multiple funding rounds, including a $100 million Series F in 2020 and a $110 million Series G in 2021 which set a valuation of $2 billion. By 2017, the database appeared in the top 5 most loved databases in Stack Overflow developer surveys.

The license pivot and the community split

The March 2024 license change from BSD-3-Clause to a dual SSPL/RSALv2 model split the community. This shift led to the creation of Valkey, a fork that Amazon and Google support on their managed services. In late 2024, Sanfilippo rejoined the company to work as a developer evangelist. He currently works on vector capabilities and the integration of Redis Stack technologies into core Redis 8. Redis reached $300 million in revenue in January 2026. The company changed its name from Redis Labs to Redis in 2021 and relocated its headquarters from Mountain View to San Francisco in mid-2024. The company reached 501 employees by October 2024. Sanfilippo is implementing HNSWs from scratch to support new vector data types. Redis 8 also includes 30 performance improvements that deliver up to 87% faster commands and 2x throughput. This version also adds the OSI-approved AGPL as an additional licensing option. Will the current community momentum stay with Redis or shift toward Valkey? You should analyze your own dependency on the RESP protocol before selecting a specific fork.

Choosing between KeyDB and Dragonfly

In 2026, engineers choose between Redis, Valkey, KeyDB, and Dragonfly based on specific workload needs. Engineers testing on an Apple M4 chip with 8 CPU cores and 4 GiB of memory find that Dragonfly achieves 5x higher throughput than Redis on large multi-core instances because it uses a shared-nothing multi-threaded architecture called "dashtable." It also uses 20-30% less memory for small objects, which can reduce infrastructure costs by 20-40% for workloads with millions of small objects. Dragonfly shows a 125% improvement in scalability when moving from 1 to 4 threads. Redis 8.4.0 provides the lowest latency for simple request patterns and reaches 16,605 ops/sec in PHP-based WordPress object cache tests. Valkey reaches 16,561 ops/sec in those same tests and leads in batched request throughput. KeyDB holds the title for the single-thread champion with 66,443 ops/s. KeyDB also shows the most resilience to TLS, which causes a 30.3% throughput drop and a 40% latency increase in other engines. In the same PHP tests, KeyDB shows 15-18% lower throughput than Redis or Valkey. In the pub/sub fanout test, Valkey shows higher p95 latency than Redis and KeyDB. When evaluating the 1:15 SET:GET ratio with 512 Byte payloads, Redis 8.4.0 reaches 125,524 ops/sec while Dragonfly v1.37.0 reaches 119,615 ops/sec.