The history of the Redis and Valkey split

Salvatore Sanfilippo created Redis in 2009 to solve MySQL performance issues for his LLOOGG analytics service. His design used in-memory storage and atomic operations to manage data like lists and sorted sets. Sanfilippo stepped down as maintainer in 2020 to focus on his family. In March 2024, Redis Ltd. changed the license from BSD to a dual SSPLv1 and RSALv2 model. This change stopped cloud providers from using the source code for free.

The community created Valkey.

Origins and the license shift

The licensing change split the developer community. Former maintainers and companies including AWS, Google Cloud, Oracle, Ericsson, and Snap launched Valkey under the Linux Foundation in March 2024. Madelyn Olson, a former Redis maintainer and current AWS principal software engineer, led the effort to build a new open community. This project remains available under the BSD license.

Which engine will dominate the next decade?

Valkey and the Linux Foundation

Valkey is a high-performance key/value datastore. It supports caching, message queues, and serving as a primary database. Madelyn Olson, a principal software engineer at AWS, noted the community’s rapid growth. The project reached 500,000 container pulls. It also has 1,000 commits.

Valkey 8.0 improved memory efficiency by using a dictionary per slot and embedding keys into dictionary entries. This method reduced memory overhead by 8 bytes per key and allowed for 20% more capacity. Valkey 9.0 introduced asynchronous I/O threading and SIMD acceleration for AVX512 and NEON. AWS reports that Valkey 9.0 provides up to 40% higher throughput for pipelined workloads.

Valkey 8.1 arrived in April 2025. The project now has support from over 40 organizations. Percona research shows that 70% of respondents with Redis deployments want alternatives because of the license change. AWS provides Amazon MemoryDB for Valkey, which offers a lower price point than existing Redis versions. For Azure users, Microsoft plans to retire Azure Cache for Redis tiers by 2027 and 2028, making migration necessary.

Valkey 9.0 is fast.

Divergent engines and feature sets

The divergence between the two platforms is real. Redis 8.0 integrated Redis Stack modules like JSON, Time Series, and the Query Engine into the core in May 2025. This version also introduced native vector sets. Sanfilippo returned to the company in November 2024 as an evangelist. He contacted the CEO, Rowan Trollope, to discuss his return. He is working on a design document to implement HNSWs from scratch. His work targets vector search for AI.

Redis 8.8 added batched prefetch for MGET, MSET, and HGETALL. Redis 8.0 delivers up to 87 percent faster commands and double the throughput.

Valkey provides modules like valkey-json, valkey-bloom, and valkey-ldap as separate tools. However, valkey-search is a subset that lacks autocomplete, synonyms, spellcheck, index aliasing, query profiling, cursors, and GEOSHAPE polygon queries. Time Series has no equivalent in Valkey.

The decision between these engines is a matter of licensing and specific module needs. If your organization prohibits AGPL-licensed dependencies, choose Valkey. If you depend on Time Series or advanced search features, stay with Redis. Redisson supports both engines, meaning the API remains the same for Java developers regardless of the engine choice. You should check your module dependencies before switching.