In March 2024, Redis Ltd. changed the Redis license from BSD to the Server Side Public License (SSPL) - the same move MongoDB made in 2018. The community response was swift: within a week, the Linux Foundation announced Valkey, a community fork of Redis under the original BSD license. AWS, Google Cloud, Oracle, and Ericsson all joined as founding members.
Two years later, both projects have shipped significant versions. Redis 8 launched with substantial performance improvements. Valkey 8 launched with its own performance claims and full compatibility with the Redis API. The question everyone is asking: which one should you actually use?
What Changed and Why It Matters
The SSPL is not open source by the OSI definition. The short version: if you use Redis as part of a service you provide to others, you must either open-source your entire service stack or get a commercial license from Redis Ltd.
For most developers building applications that use Redis internally, this doesn’t affect you. You’re a user, not a service provider.
For cloud providers (AWS, Google, Azure) or anyone building a Redis-as-a-service product, the SSPL is a dealbreaker. This is why every major cloud provider pivoted to offering Valkey instead of Redis for new managed service instances.
Performance: What Both Projects Claim
Redis 8 highlights:
- 87 new commands added
- Hash field expiry (finally)
- Improved memory efficiency for small objects
- Vector set data type (Redis is pushing into AI use cases)
Valkey 8 highlights:
- I/O threading improvements: reads and writes can happen in parallel
- Claimed 46% throughput improvement over Redis 7.2 on multi-core systems
- Full API compatibility with Redis 7.2
Independent benchmarks from Percona (Q4 2025) showed:
| Workload | Redis 8 | Valkey 8 |
|---|---|---|
| GET throughput (ops/sec) | 680,000 | 890,000 |
| SET throughput (ops/sec) | 650,000 | 850,000 |
| Mixed GET/SET | 670,000 | 870,000 |
| Latency p50 (GET) | 0.3ms | 0.2ms |
| Latency p99 (GET) | 1.2ms | 0.9ms |
The Valkey performance edge is real and comes from its multi-threaded I/O implementation. Redis has historically been single-threaded for command processing (with threaded I/O added in Redis 6), and Valkey pushed this further.
API Compatibility
If you’re migrating from Redis to Valkey, the answer is: almost perfectly compatible.
Valkey targets 100% API compatibility with Redis 7.2. Every Redis client library works with Valkey without changes. The Valkey team’s stated policy is that any Redis 7.2-compatible command will work identically in Valkey.
Redis 8 added new commands that Valkey doesn’t have yet, and vice versa. The EXPIREAT hash field expiry in Redis 8 is genuinely useful. Valkey has its own additions. If you’re building new features targeting specific Redis 8 commands, you’d need to check Valkey’s roadmap.
For existing applications using Redis as a cache or session store, migration is copy the config, swap the endpoint.
Ecosystem and Cloud Support
This is where Valkey has a decisive advantage:
| Provider | Redis offering | Valkey offering |
|---|---|---|
| AWS | ElastiCache Redis (existing) | ElastiCache Valkey (default for new) |
| Google Cloud | Memorystore Redis | Memorystore Valkey |
| Azure | Azure Cache for Redis | Azure Cache for Redis (Redis) |
| Aiven | Aiven Redis | Aiven Valkey |
| Upstash | Upstash Redis | Upstash Valkey |
AWS made Valkey the default for new ElastiCache instances in mid-2024. Google followed. Azure is the notable holdout - Microsoft has a deeper commercial relationship with Redis Ltd.
If you’re deploying on AWS or GCP, Valkey is the path of least resistance for new projects. The managed services are better-supported and positioned as the go-forward option.
Redis’s Actual Advantage
Redis Ltd. has engineering resources and commercial incentives to ship features that pure community projects move slower on. The vector set data type in Redis 8 and the JSON module (RedisJSON) are feature-rich and well-maintained.
If you need RedisSearch, RedisTimeSeries, or RedisJSON - the Redis Stack modules - there are Valkey-compatible alternatives in development, but the Redis Ltd. versions have more polish and documentation today.
Redis Enterprise (the paid offering) has genuine enterprise features: active-active geo-distribution, Redis on Flash (tiered storage), and dedicated support. If you’re at the scale where these matter, you were already paying for Redis Enterprise anyway and the license change may not affect your evaluation.
Which One Should You Use
For new projects on AWS or GCP: Valkey. The cloud integration is better, it’s genuinely faster, and the license is cleaner. Drop-in compatible with everything Redis.
For existing Redis deployments: evaluate migration cost vs. benefit. If you’re on managed Redis (ElastiCache, Cloud Memorystore), check whether a Valkey migration path is available. AWS has tooling for this. If you’re self-hosting, Valkey is a straightforward fork and the API compatibility means low risk.
For Redis Stack features (search, JSON, time series): Redis has the better story today. Valkey’s ecosystem is catching up, but if you’re depending heavily on Redis modules, check the specific module compatibility before migrating.
For on-premise with support contracts: Redis Enterprise is still the enterprise choice. Valkey support options are less mature.
Bottom Line
Valkey won the open-source fork race. It has major cloud provider backing, better performance, a cleaner license, and a strong community that includes Redis’s original contributors. For most applications that use Redis as a cache, queue, or session store, Valkey is the better choice in 2026 and the default on major cloud platforms.
Redis retains advantages in its paid tiers and proprietary module ecosystem. The community Redis vs. Valkey question has been settled - Valkey. The enterprise features question is more nuanced.
Comments