Maintainers of the open source database Valkey announced Valkey 9.1 this week at Open Source Summit North America in Minneapolis. Now generally available, the latest release adds security and observability features, performance upgrades and memory-efficiency improvements.
On the security side, Valkey 9.1 adds database-level ACLs, allowing administrators to restrict users to specific numbered databases. The release also moves the Lua scripting engine into a separate module, reducing Valkey’s attack surface by letting administrators disable it when it is not required. TLS updates add certificate expiration visibility, automatic background certificate reloading, and support for SAN URI-based mTLS authentication to help avoid outages and downtime.
Valkey 9.1 also adds observability improvements for teams monitoring Valkey in production. New main and I/O-thread usage metrics can give a clearer view of actual server load, addressing cases where CPU utilization can appear high when Valkey threads are waiting for work in a busy loop. The release also introduces JSON logging to make server logs easier to ingest into observability platforms without custom parsing.
The new release also includes several performance improvements, with the project reporting single-server throughput of up to 2.1 million requests per second under specific benchmark conditions. The release adds a redesigned I/O threading communication model that developers say improves throughput by up to 17% across a range of workloads, along with faster stream range operations and up to 30% higher throughput for some string GET commands.
Valkey 9.1 also improves sorted set queries, caches COMMAND responses to reduce connection setup time for some clients, and enables hardware clock use by default, which developers say can reduce the overhead of time-related system calls and improve GET and SET performance by up to 3%.
The new version also continues Valkey’s focus on memory efficiency, with internal pointer optimizations that can reduce memory usage by up to 20% for strings under 128 bytes and skiplist optimizations that reduce sorted set memory usage by up to 10%. Valkey 9.1 also improves internal operations tied to latency and resource use, including optimized hash table rehashing, faster bulk deletes, and more efficient replica creation when AOF is enabled.
Madelyn Olson, a Valkey project maintainer and principal engineer for in-memory databases at AWS, said in a summit keynote that her team used the 9.1 release cycle to reduce some of the repetitive work involved in shipping releases. That includes AI-assisted tooling for backporting commits from the development branch to release branches, with tests and human review still built into the process.
That work comes as Valkey maintainers are handling a sharp rise in AI-assisted contributions. Olson said the project has seen a 55% increase in individual commits over the past six months and a roughly 500% increase in lines of code submitted to maintainers, adding to the amount of code they must review. While AI agents can help with that workload, Olson said they should be used to support maintainers rather than add more pressure or shift project responsibility away from the people who understand the codebase and its users.
“We really want to make sure that these agents are being used to support the existing maintainers of the project,” she said. “The maintainers are still human, and I’ve seen so much more burnout in the age of AI than I have in quite a while.”
