Lux vs Redis

Lux stays Redis-compatible where it matters, but it is not trying to be just another Redis host. It is a broader data runtime with tables, vectors, queues, time series, and Lux Cloud on top.

Benchmarks below use redis-benchmark against Redis 8.6.1 and Lux on identical hardware with identical parameters. Treat them as workload-specific measurements, not universal claims.

Throughput (requests/sec)

SET Pipeline 64
3.4x
Lux
11,235,955
Redis
3,300,330
GET Pipeline 64
2.6x
Lux
12,048,193
Redis
4,694,836
INCR Pipeline 64
1.6x
Lux
6,289,309
Redis
4,032,258
LPUSH Pipeline 64
2.0x
Lux
6,493,507
Redis
3,322,259
RPUSH Pipeline 64
1.7x
Lux
6,410,257
Redis
3,690,037
LPOP Pipeline 64
3.9x
Lux
11,627,907
Redis
3,012,048
RPOP Pipeline 64
3.4x
Lux
11,111,111
Redis
3,257,329
SADD Pipeline 64
1.8x
Lux
7,194,245
Redis
4,081,633
HSET Pipeline 64
2.0x
Lux
6,756,757
Redis
3,311,259
SPOP Pipeline 64
2.7x
Lux
12,195,122
Redis
4,484,305
ZADD Pipeline 64
2.3x
Lux
6,993,007
Redis
3,105,590
ZPOPMIN Pipeline 64
2.2x
Lux
11,494,253
Redis
5,319,149

Product Comparison

FeatureLuxRedis
LanguageRustC
ArchitectureMulti-threaded, shardedSingle-threaded
ScalingScales with CPU coresNeeds clustering
Peak measured throughput12M ops/sec on this benchmark~5.3M ops/sec on this benchmark
StreamsYesYes
Lua scriptingYes (Lua 5.4)Yes (Lua 5.1)
BullMQYesYes
Blocking commandsBLPOP, BRPOP, BZPOPMINFull support
Vector searchBuilt-in (VSET/VSEARCH)Requires Redis Stack
TablesBuilt-inNot built-in
Realtime subscriptionsBuilt-inRaw pub/sub only
Time seriesBuilt-inRequires Redis Stack
Binary size~2MB~8MB
ProtocolRESP2RESP2 + RESP3
Client librariesAll Redis clients workNative
LicenseMITRSALv2 / SSPL
Managed hostingLux Cloud ($10/mo per project)Redis Cloud ($$$)

Where Lux fits

Use Lux when you want Redis-compatible performance and command coverage, but you also want your app data stack to stay in one system. That means cache, tables, vectors, realtime, jobs, and time series living together instead of being split across separate products.

Why Lux is faster

Redis is single-threaded by design. Lux uses Rust and sharded multi-core execution, so throughput scales with available CPU instead of hitting one-core limits early. Pipeline batching and zero-copy protocol handling reduce overhead on the hot path.

How we benchmark

We use redis-benchmark against both Lux and Redis on identical hardware with identical parameters. The headline numbers are strongest for pipelined single-key workloads. Multi-key commands, large values, persistence settings, and network placement can change the result, so serious evaluations should run the benchmark mix that matches the production workload.

When Redis is still the better fit

Use Redis if you specifically need Redis Cluster, RESP3, modules, or ACLs today. Lux covers a large and growing set of production workloads, but the reason to switch is not just benchmark speed. It is the combination of Redis compatibility with a broader runtime for application data.