Sseekdeploy
Submit a tool

Valkey

BSD-licensed, Redis-compatible in-memory data store: the Linux Foundation fork of Redis 7.2.

Open source · BSD-3-ClauseFree#key-value#cache#in-memory#redis-compatible#open-source
Visit website ↗DocumentationGitHub

What is Valkey?

Valkey is an in-memory key-value store: one server process holds the dataset in RAM and speaks the Redis protocol (RESP), so existing Redis clients and most commands work unchanged. It keeps Redis's data structures, RDB snapshots and append-only file, replication, Sentinel and cluster mode. JSON, Bloom filters, search and LDAP auth are separate modules, shipped together in the valkey/valkey-bundle image.

Valkey began in March 2024 as a fork of Redis 7.2.4, after Redis left the BSD licence. It's a Linux Foundation project under BSD-3-Clause, developed by maintainers from several companies with no single owner selling it. AWS (ElastiCache), Google Cloud (Memorystore) and DigitalOcean sell managed Valkey.

The current release is 9.1.3 (26 October 2026, a security fix), and 9.2 is in release candidates (rc2 on 4 November 2026). Valkey 9.0 (October 2025) added hash-field expiry, atomic slot migration and multiple databases in cluster mode. Redis itself, now available under AGPLv3, is the main alternative.

Our review

4.5

The safe default for a new Redis-shaped cache, queue or session store, unless you need Redis 8's built-in extras. Valkey speaks the same protocol, works with the same clients and keeps shipping real improvements (a vendor's reproducible sorted-set benchmark has 9.1 using about 10% less memory than Redis 8.10), all under a BSD licence nobody needs to read twice. It's also the cheaper engine on the big clouds: ElastiCache charges 20% less per node and 33% less on Serverless than for Redis OSS. What you give up is Redis 8's all-in-one core: JSON, search and Bloom filters are separate Valkey modules, and some newer Redis commands aren't there. Budget for patching too: three of the 9.1 releases carried security fixes in six months. Who gets paged is the same question as with Redis; Valkey doesn't make running it yourself any easier.

Pros

  • + BSD-3-Clause under the Linux Foundation, so embedding it or selling it as a service needs no licence review
  • + Drop-in for Redis 7.2 clients and commands, with active performance work: in Momento's September 2026 benchmark, 3.34 GB for 50M sorted-set members against Redis 8.10's 3.73 GB
  • + Cheapest managed engine on AWS (Serverless from about $6 a month, nodes 20% under Redis OSS), and also offered by Google Cloud and DigitalOcean

Cons

  • − Redis 8's core JSON, query engine, time series and vector sets aren't in the Valkey server; you add modules or do without
  • − Frequent security releases: 9.1.0, 9.1.2 and 9.1.3 (May to October 2026) all fixed vulnerabilities, one of them reachable without authentication
  • − Cluster mode still needs cluster-aware clients that handle MOVED/ASK redirects and CROSSSLOT errors

Reviewed by SeekDeploy Editorial · updated 2026-11-15. Reviews are independent — we don’t accept payment for coverage or ratings.

Should you use Valkey?

A good fit if you're…

  • New caches, session stores, rate limiters and job queues built on Redis clients, where a BSD licence ends the legal conversation
  • Teams on AWS or Google Cloud choosing a managed cache engine: Valkey is the cheaper option on ElastiCache and the engine behind Memorystore for Valkey
  • Companies that embed or resell an in-memory store and can't take on AGPLv3, SSPL or RSALv2

Probably not for you if you're…

  • Apps built on Redis 8's built-in JSON, query engine, time series or vector sets that want them in one server with Redis's commands
  • Teams that want a single vendor behind the project for support contracts: Valkey support comes from cloud providers and third parties
  • A tiny queue or cache next to an app that already has Postgres, where another service to run and patch isn't worth it

Pick something else if…

  • you want Redis 8's JSON, query engine, time series and vector sets in the core server, or Redis Cloud's active-active, and AGPLv3 is acceptable → Redis
  • your queue or cache is small and you'd rather not run and patch another service → PostgreSQL

Gotchas to know before you commit

  • Security releases come often: 9.1.0 (May 2026) fixed three CVEs, 9.1.2 (August) an unauthenticated use-after-free in the Lua debugger, and 9.1.3 (October) a heap overflow in cluster Pub/Sub. Track the release notes and patch promptly.
  • Before 9.1.3, Valkey could busy-spin the CPU on virtual machines whose kernel clocksource isn't tsc, such as kvm-clock on AWS (fixed in 9.1.3, #4691). Sentinel spinning at 100% CPU after boot was also reported (#4766, closed as of 2026-11-15).
  • Docker tags: valkey/valkey:9.2 was pushed as a 9.2.0 release candidate (#4841, closed as of 2026-11-15). RCs now go to 9.2-rc, but a 9.2 tag from 21 September 2026 is still on Docker Hub, so pin 9.1 or an exact version until 9.2 is GA.
  • It tracks Redis 7.2 compatibility, not Redis 8: commands added in Redis 8 may be missing or arrive under different names, and JSON, search and Bloom need the modules (valkey-bundle).
  • Cluster mode needs clients that follow MOVED/ASK redirects, and multi-key commands across hash slots fail with CROSSSLOT; the proposal for a valkey-proxy to hide this (#4761) was closed as completed on 2026-11-02, so check the release notes for whether a proxy has shipped before relying on one.

Valkey pricing

Checked Nov 15, 2026 against the official pricing page

Valkey itself is free (BSD-3-Clause); what you pay for is RAM, on your own server or a managed service. Managed Valkey starts around $6 a month on ElastiCache Serverless, but Serverless storage is $0.084 per GB-hour (about $61 per GB-month), so beyond a few GB node-based clusters or your own servers are far cheaper.

Open source
$0
Self-hosting
  • Server, Sentinel, cluster mode, RDB and AOF persistence
  • Modules (JSON, Bloom, search, LDAP) in valkey/valkey-bundle
  • Limit: You run failover, backups, upgrades and security patches
AWS ElastiCache Serverless
$0.084 / GB-hour + $0.0023 / million ECPUs
Spiky or small caches on AWS
  • No nodes to size; scales storage and throughput
  • 33% below Redis OSS Serverless ($0.125 / GB-hour)
  • Limit: Minimum 100 MB metered per cache (~$6.13 / month)
  • Limit: 1 ECPU per KB transferred; storage cost grows fast past a few GB
AWS ElastiCache nodes
From $0.0128 / hour
Steady caches on AWS
  • cache.t4g.micro (0.5 GiB) $0.0128/h; r7g.large (13.07 GiB) $0.1752/h
  • 20% below the Redis OSS node price
  • Limit: You size and pay for each node, replicas included
  • Limit: us-east-1 on-demand prices
Google Memorystore for Valkey
From $0.0318 / node-hour
Teams on Google Cloud
  • shared-core-nano 1.4 GB $0.0318/h; standard-small 6.5 GB $0.1425/h; highmem-medium 13 GB $0.1923/h
  • Committed-use discounts: 20% (1 year), 40% (3 years)
  • Limit: Billed per node, so each replica is another node
  • Limit: us-central1 prices; backups and AOF cost extra

What it costs in practice

Cache on the app's VPS
$0 extra
Valkey from the distro, Homebrew or the valkey/valkey image on a server you already pay for, with a few hundred MB of RAM to spare. Nobody else is on call: you own restarts, backups and patching.
Small SaaS: sessions and queue
~$15–37 / month
About 500 MB of data, ~50 small requests a second. ElastiCache Serverless: 0.5 GB × $0.084 × 730 h = $30.66 + 131.4M ECPUs × $0.0023/M = $0.30, so $30.96. Node-based with a replica: 2 × cache.t4g.small (1.37 GiB) × $0.0256 × 730 = $37.38. DigitalOcean Managed Caching for Valkey: 1 GiB for $15 a month, single node.
10 GB working set with a replica
~$256–613 / month
ElastiCache nodes: 2 × r7g.large (13.07 GiB) × $0.1752 × 730 = $255.79. Memorystore: 2 × highmem-medium (13 GB) × $0.1923 × 730 = $280.76. ElastiCache Serverless: 10 GB × $0.084 × 730 = $613.20 before ECPUs. Serverless only wins when the data is small or traffic is very spiky.

Estimates in USD from list prices. Your bill depends on usage, region, and discounts.

Health & activity

Updated Nov 15, 2026 · last 90 days
Actively developedFrequent commits or releases in the last 90 days.
GitHub stars
28k
1.5k forks
Commits
375
at least 41 committers
Merged pull requests
464
Issues opened / closed
219 / 190
median 48.6 days to close
Last push
Nov 15, 2026
Open issues & PRs
1k

From valkey-io/valkey. Counts cover the last 90 days unless noted.

Downloads

Benchmarks

50M-member sorted set: memory (Momento)

Source: Momento, "50 Million Sorted Sets, Round Three: Redis Still Chasing Valkey" (27 September 2026)

Server used_memory after inserting 50 million members into one sorted set with pipelined ZADD NX · GB, lower is better

  • Valkey 9.1.23.34 GB
  • Redis 8.10.23.73 GB
  • Valkey 9.0.63.77 GB
  • Redis 8.4.74.83 GB
  • Valkey 8.0.114.83 GB
How this was measured

momentohq/sorted-set-benchmark (public): 50M members m:{i} with integer scores into a single ZSET via pipelined ZADD NX; official Docker images with --save '' --appendonly no; one server at a time and one client on the same r8g.4xlarge (Graviton4); ten repetitions per version with a flush between. Every repetition reported the same used_memory at ~10 MB precision. Median inserts per second: Valkey 9.1.2 647k, Redis 8.10.2 632k. Not neutral: Momento sells a Valkey-based cache service, and this measures one data structure on one machine, not general throughput. Our own Redis vs Valkey script (benchmarks/redis/) hasn't been run. Five of 11 versions shown; copied 2026-11-15.

Versions: Valkey 9.1.2, Redis 8.10.2

Published Sep 27, 2026 · source

What the community says

Hacker News & GitHub issues · Aug 17, 2026 – Nov 15, 2026

On Hacker News, Valkey mostly comes up as the example of a community fork that worked after a licence change, and as the Redis replacement companies and self-hosted stacks now ship. On GitHub the conversation is about the 9.2 cycle: new data types, cluster tooling, module APIs and a handful of CPU and migration bugs.

What people like

  • The fork that stuck · 4 sources

    Commenters cite the move from Redis to Valkey as the model outcome of a licence change: companies switched their internal caches, outside contributors went to Valkey, and it's named next to OpenTofu as a counterexample to open-source rug pulls.

    HN comment 1HN comment 2HN comment 3
  • Cloud vendors pay for it · 2 sources

    People note that AWS supports Valkey and pays some of its maintainers, and read that backing as why the fork has staying power.

    HN comment 1HN comment 2
  • Performance and memory work keeps landing · 2 sources

    Shared articles cover Valkey 9.2 cutting memory overhead and a third round of sorted-set benchmarks framed as Redis catching up with Valkey.

    HN thread 1HN thread 2

What people complain about

  • Cluster mode pushes work onto clients · 2 sources

    The most-engaged issue in the window proposes a valkey-proxy because clients must track slot maps, handle MOVED/ASK redirects and work around CROSSSLOT errors; another reports a failed slot migration leaking a child process that blocks all further migrations on that node.

    GitHub issue 1GitHub issue 2
  • CPU spinning on virtual machines · 2 sources

    Users report Sentinel spinning at 100% CPU after boot and the event loop spinning on AWS instances that use kvm-clock after upgrading to 9.1.2; 9.1.3 ships a fix for the TSC clock case.

    GitHub issue 1GitHub issue 2
  • Gaps in the module API · 3 sources

    Module authors report low-level key APIs that skip keyspace notifications (which desyncs search), blocking calls that silently return NULL for RESP3 Pub/Sub clients, and a missing aligned allocator.

    GitHub issue 1GitHub issue 2GitHub issue 3

Written by Petra Lindqvist on Nov 15, 2026 after reading 35 relevant Hacker News comments, threads and GitHub issues from Aug 17, 2026 to Nov 15, 2026. Each point links to the discussions behind it, and the count shows how many sources raised it. This is a reading of public discussion, not a survey.

On Hacker News · 4 threads, 19 comments mentioning it

Alternatives to Valkey

Redis

Databases & Backends
4.5

In-memory data store for caching, queues, sessions, and real-time workloads.

Open sourceFreeActiveReviewed

PostgreSQL

Databases & Backends
5.0

The world's most advanced open-source database — and the industry default.

Open sourceFreeActiveReviewed

Spotted something wrong or out of date? Flag a correction on GitHub. Health and community data refresh on the 1st and 15th of every month.