Valkey vs Redis in 2026: Migrating After the License Change
Valkey vs Redis in 2026: how the fork happened, real benchmark differences, drop-in compatibility, and whether migrating off Redis is worth it for your stack.
In March 2024, Redis Inc. changed its license away from open source, and within a week a Linux Foundation–backed fork called Valkey appeared with the original Redis maintainers and cloud heavyweights behind it. Two years later the question every backend team asks during an infra review is the same: Valkey vs Redis in 2026 — is the fork a safe drop-in, and is there any reason to stay on Redis?
If you provisioned a managed cache recently, you may have noticed the default quietly shifted. AWS, Google Cloud, and Oracle all offer Valkey now, and in several cases it’s cheaper than the Redis-branded option for the same instance class. That pricing gap alone is pushing teams to take the migration seriously.
TL;DR
- Valkey is a fork of Redis 7.2, governed by the Linux Foundation under the permissive BSD-3 license. It’s still open source; Redis’s main offering moved to the source-available RSALv2/SSPLv1 dual license (Redis later re-added an AGPL option in v8).
- Wire-protocol and command compatible — existing Redis clients,
redis-cli, and most tooling work unchanged against Valkey. - Valkey 8.x added multi-threaded I/O improvements that push throughput meaningfully higher than Redis 7.2 on multi-core nodes.
- For the vast majority of cache/queue workloads, migration is a connection-string change, not a rewrite.
How the Fork Happened
Redis spent over a decade as the default open-source in-memory data store under a BSD license. When Redis Inc. relicensed the core to RSALv2 and SSPLv1 — both “source available” but not OSI-approved open source — the goal was to stop cloud providers from selling managed Redis without contributing back.
The community response was immediate. The Linux Foundation launched Valkey, forked from the last BSD-licensed release (Redis 7.2.4), and the original lead Redis contributors moved over. AWS, Google, Oracle, and Ericsson committed engineering resources. Redis Inc. responded in 2025 by adding an AGPLv3 option to Redis 8, restoring an open-source path — but by then Valkey had already become the default in much of the managed-cloud world.
This matters for one practical reason: if you build a commercial product on top of Redis under SSPL, you inherit licensing obligations around offering it “as a service.” Valkey’s BSD-3 license has none of that ambiguity, which is why it shows up in backend system design reviews as the lower-risk default.
Performance: What Actually Changed
For a long time the honest answer was “they’re identical” — Valkey 7.2 was a byte-for-byte behavioral fork. That changed with Valkey 8.0, which invested in asynchronous I/O threading and per-slot dictionary improvements that Redis 7.2 doesn’t have.
The numbers below are representative of single-node memtier_benchmark runs on an 8-vCPU instance with a 50/50 GET/SET mix. Your results depend heavily on payload size, pipelining, and network path — always benchmark your own workload before committing.
| Metric | Redis 7.2 | Valkey 8.x | Notes |
|---|---|---|---|
| Single-thread GET/SET | Baseline | ~Same | Core engine is shared lineage |
| Multi-core throughput | Baseline | ~1.2–2.3× higher | Valkey’s enhanced I/O threading |
| p99 latency under load | Baseline | Lower on busy nodes | Less main-thread contention |
| Memory per key | Baseline | Slightly lower (8.1+) | Embedded key optimizations |
| License | RSALv2 / SSPL / AGPL | BSD-3 | OSI open source |
The headline: on a single core they’re effectively tied, but on the multi-vCPU instances most teams actually run, Valkey 8’s threading model pulls ahead. If your cache is CPU-bound at peak — common when you’ve layered in smart caching strategies like write-through and request coalescing — that headroom is real.
Where they’re still the same
Data types, persistence (RDB/AOF), replication, Lua scripting, Pub/Sub, and clustering all behave identically. Valkey reads existing RDB and AOF files, so a migration doesn’t require a data export — you point a replica at your primary and promote it.
Compatibility and Migration Path
Because Valkey forked from a released Redis version, client compatibility is essentially complete for anything built on Redis 7.2 and earlier semantics. ioredis, redis-py, go-redis, lettuce, and node-redis all connect to Valkey with no code changes.
The cleanest migration uses replication:
- Stand up a Valkey replica of your existing Redis primary (
REPLICAOF your-redis-host 6379). - Let it sync fully and confirm
INFO replicationshows it caught up. - During a low-traffic window, promote the replica and repoint your app’s connection string.
- Keep the old Redis node warm for a quick rollback until you’ve watched a full traffic cycle.
For teams already running infrastructure as code, this is usually a one-line provider or engine change — similar in effort to swapping a serverless compute backend. The risk surface is small precisely because the protocol is unchanged.
Pros and Cons
| Valkey | Redis (8.x) | |
|---|---|---|
| License | BSD-3, fully open source | AGPL or commercial RSALv2/SSPL |
| Multi-core throughput | Higher (8.x threading) | Strong, improving |
| Managed cloud support | AWS, GCP, Oracle — often cheaper | Available, sometimes premium-priced |
| New features | Community-governed roadmap | Vector sets, Redis-specific modules first |
| Ecosystem maturity | Younger, fast-growing | Decade-plus, largest install base |
| Module ecosystem | Some Redis modules are not relicensed | Full first-party module suite |
Who Should Migrate (and Who Shouldn’t)
Migrate to Valkey if: you use Redis as a cache, session store, rate limiter, or queue; you want guaranteed open-source licensing; or your cloud bill drops by choosing the Valkey-branded managed tier. This covers most production workloads — including the kind of token-bucket rate limiting that lives in a tiny set of standard commands.
Stay on Redis if: you depend on Redis-specific modules (RedisJSON, RediSearch, vector sets) that aren’t available or fully matched in Valkey, or you’ve bought commercial Redis Enterprise support and value the SLA.
Either is fine if: you’re a small project where the cache is a few hundred MB and never CPU-bound — at that scale the difference is academic, much like the SQLite vs Postgres trade-off where either choice works until you actually have scale problems.
Frequently Asked Questions
Is Valkey a drop-in replacement for Redis?
For Redis 7.2-era features, yes. Same wire protocol, same commands, same client libraries, and it reads existing RDB/AOF files. Modules and post-7.2 Redis-exclusive features are the main exceptions.
Is Valkey actually faster than Redis?
On a single core they’re nearly identical. On multi-core nodes, Valkey 8’s improved I/O threading gives it a throughput edge. Benchmark your own payload and access pattern before assuming a number.
Will my Redis client library work with Valkey?
Yes. ioredis, redis-py, go-redis, lettuce, and others connect to Valkey without code changes because the protocol is unchanged.
Is Redis no longer open source?
Redis 8 re-added an AGPLv3 option, so an open-source path exists again. But the commercial offering still uses source-available licenses, which is why many teams prefer Valkey’s straightforward BSD-3.
Can I run Valkey and Redis side by side during migration?
Yes — set up a Valkey replica of your Redis primary, sync, then promote and repoint. Keep the old node warm for rollback.
Bottom Line
For typical caching, queue, and session workloads, Valkey is the safer 2026 default — open-source license, drop-in compatibility, faster on multi-core, and often cheaper on managed cloud. Stay on Redis only if you depend on its proprietary modules or commercial support.
Recommendations are based on independent research and testing. Always validate against your own workload before migrating production systems.
Planning a broader caching overhaul? Start with our backend caching strategies guide, then revisit your system design fundamentals before you cut over.
Sources: Valkey official site · Valkey GitHub · Redis licensing announcement