Why Valkey’s AWS-backed fork matters for Redis users

The licensing fallout and the AWS response

I recall the industry disruption in March 2024 when Redis transitioned from the BSD license to the SSPLv1 and RSALv2. This transition ended the open-source status of Redis and forced cloud providers to find alternatives. Valkey emerged as a CNCF-backed fork of Redis 7.2.4, maintaining the permissive BSD 3-clause license. AWS, Google Cloud, Oracle, Ericsson, and Snap supported the project. Madelyn Olson, a principal engineer at Amazon ElastiCache and former Redis contributor, helped lead the fork.

Valkey is the path.

In May 2025, Redis 8 returned to being open source by adding the AGPLv3 license as an option. This change followed the return of creator Salvatore Sanfilippo to the project. However, the AGPLv3 license carries a strong copyleft requirement. If you modify AGPL-licensed code and run it as a network service, you might need to disclose those modifications. You know the drill. This reality maintains Valkey’s relevance for businesses that prioritize permissive licensing. Redis 7.2 reached end-of-life on February 28, 2026.

Divergent development in performance and features

The two projects have diverged since the fork. Redis 8.6 now includes the previously separate Redis Stack modules like JSON and Time Series in its core distribution, which simplifies the experience for developers building AI-driven applications. This integration includes native support for vector sets and the Redis Query Engine. Redis 8.0 reported up to 87% faster command execution and 2x higher throughput compared to earlier versions. Redis 8.4 also added hybrid search using the FT.HYBRID command.

Valkey focuses on core efficiency.

Valkey 8.0 introduced asynchronous I/O threading and a dictionary-per-slot design that reduces per-key metadata overhead. This architectural change cuts metadata overhead by several dozen bytes per key. Valkey 9.0 added aggressive SIMD acceleration and pipeline memory prefetching to improve throughput. AWS reports that Valkey 9.0 delivers up to 40% higher throughput for pipelined workloads. Valkey’s RDMA support remains experimental and is limited to Linux environments without TLS support for replication.

The gap in modules remains real. Valkey provides the valkey-json and valkey-bloom modules, but it lacks the maturity found in the Redis core. If your workload depends on RedisTimeSeries or advanced search functions like relevance ranking, the current lack of mature Valkey equivalents makes switching a significant engineering risk. Valkey’s search module is a subset that misses autocomplete, synonyms, and query profiling.

The decision for the user

The choice depends on your infrastructure and specific data needs. AWS offers ElastiCache for Valkey, and it costs roughly 20% less than ElastiCache for Redis. For session caching, rate limiting, or leaderboard functions, Valkey is the logical choice.

The choice remains difficult.

I see many teams running into client library issues during migrations. Some libraries struggle with version negotiation between Valkey 8.x and what they expect from Redis 7.x. You must audit your client compatibility before you attempt an upgrade. If you rely on the Redis Query Engine or specialized vector capabilities, you should stay on Redis. The migration for the global travel technology leader succeeded because they chose an in-place upgrade in ElastiCache to preserve live in-memory data without incurring any downtime for their millions of daily users.

Microsoft is retiring Azure Cache for Redis, including the Enterprise and Enterprise Flash tiers on March 31, 2027. This change forces Azure users to migrate regardless of their chosen engine.

Will the Valkey module ecosystem ever reach parity with the integrated Redis Stack?