I built a tiny, single-binary, in-memory key-value store that speaks a Redis-style RESP subset and focuses purely on speed. It’s called CQDAM Free Edition.
What it is
Single executable, zero dependencies
RESP/Redis-style protocol with pipelining
Supported: SET/GET/DEL/EXISTS, INCR/DECR, PING/INFO/HELLO, FLUSHALL
Not included (by design in Free): persistence/durability, security, clustering, observability, advanced data structures (hashes/lists/sets/zsets), pub/sub, transactions
Why
For hot paths (counters, flags, session/dedupe keys, rate limits) I wanted a smaller surface and very low overhead.
Benchmarks (how I measured)
Local loopback using redis-benchmark against CQDAM:
# Example run A (small pipeline)
redis-benchmark -h 127.0.0.1 -p 6980 -t SET,GET -n 200000 -c 10 -P 10 -d 64
# Example run B (larger pipeline)
redis-benchmark -h 127.0.0.1 -p 6980 -t SET,GET -n 500000 -c 10 -P 100 -d 64
On my laptop, run A lands ~1.2M ops/sec for SET (p50 ~0.05 ms), run B approaches the intentional Free-edition cap of ~2.5M ops/sec. Your numbers will vary by CPU, pipeline depth, payload, and NIC/loopback. I included the exact flags above so you can reproduce.
Usage
./cqdam_free 6980 &
redis-cli -p 6980 PING
redis-cli -p 6980 SET hello world
redis-cli -p 6980 GET hello
Scope & positioning
This is not a Redis replacement. It’s a minimal KV engine for ultra-hot paths where you only need a few primitives and want minimal overhead. If you need persistence, clustering, security, pub/sub, or richer data structures, use Redis (or wait for the enterprise track).
License
Free Edition is source-available, non-commercial, and intentionally performance-capped (~2.5M ops/sec). Commercial use, higher limits, and full features will be in the enterprise version.
Roadmap / ask
Enterprise Edition (targeting 7M+ ops/sec) will add durability, security, replication/cluster, and observability. I’m looking for design partners and tough workloads to validate tail latencies and failure modes.
Repo: https://github.com/LaminarInstruments/Laminar-Flow-In-Memory...
Contact: darreck@laminarinstruments.com
I built a tiny, single-binary, in-memory key-value store that speaks a Redis-style RESP subset and focuses purely on speed. It’s called CQDAM Free Edition. What it is Single executable, zero dependencies RESP/Redis-style protocol with pipelining Supported: SET/GET/DEL/EXISTS, INCR/DECR, PING/INFO/HELLO, FLUSHALL Not included (by design in Free): persistence/durability, security, clustering, observability, advanced data structures (hashes/lists/sets/zsets), pub/sub, transactions Why For hot paths (counters, flags, session/dedupe keys, rate limits) I wanted a smaller surface and very low overhead. Benchmarks (how I measured) Local loopback using redis-benchmark against CQDAM: # Example run A (small pipeline) redis-benchmark -h 127.0.0.1 -p 6980 -t SET,GET -n 200000 -c 10 -P 10 -d 64
# Example run B (larger pipeline) redis-benchmark -h 127.0.0.1 -p 6980 -t SET,GET -n 500000 -c 10 -P 100 -d 64 On my laptop, run A lands ~1.2M ops/sec for SET (p50 ~0.05 ms), run B approaches the intentional Free-edition cap of ~2.5M ops/sec. Your numbers will vary by CPU, pipeline depth, payload, and NIC/loopback. I included the exact flags above so you can reproduce. Usage ./cqdam_free 6980 & redis-cli -p 6980 PING redis-cli -p 6980 SET hello world redis-cli -p 6980 GET hello Scope & positioning This is not a Redis replacement. It’s a minimal KV engine for ultra-hot paths where you only need a few primitives and want minimal overhead. If you need persistence, clustering, security, pub/sub, or richer data structures, use Redis (or wait for the enterprise track). License Free Edition is source-available, non-commercial, and intentionally performance-capped (~2.5M ops/sec). Commercial use, higher limits, and full features will be in the enterprise version. Roadmap / ask Enterprise Edition (targeting 7M+ ops/sec) will add durability, security, replication/cluster, and observability. I’m looking for design partners and tough workloads to validate tail latencies and failure modes. Repo: https://github.com/LaminarInstruments/Laminar-Flow-In-Memory... Contact: darreck@laminarinstruments.com