tigerbeetle-benchmarks

Experimenting with 1000x Faster Financial Transaction Database built for the next 30 years. Vibe coded.

Results

$ ./tigerbeetle version
TigerBeetle version 0.16.38+2371483

$ go version
go version go1.24.2 darwin/arm64 | go version go1.24.2 linux/amd64

goos: darwin | linux
goarch: arm64 | amd64
pkg: kizzx2.com/tigerbeetle-benchmarks
cpu: Apple M4 Pro | AMD Ryzen 7 5800H with Radeon Graphics         

Benchmark Name	Mac M4 ns/op	Linux AMD ns/op	Mac M4 TigerBeetle faster by	Linux AMD TigerBeetle faster by
BenchmarkBasicTigerBeetle	24362892	3763695	-	-
BenchmarkBasicRedis	173072	537230	🔴 -140x	🔴 -6x
BenchmarkBasicPostgres	781365	2067554	🔴 -30x	🔴-1x
BenchmarkTwoPhaseTigerBeetle	60988901	16520356	-	-
BenchmarkTwoPhaseRedis	171658	560849	🔴 -354x	🔴 -28x
BenchmarkTwoPhasePostgres	1533958	7032468	🔴 -39x	🔴 -1x
BenchmarkBasicBatchTigerBeetle	27807630	30229617	-	-
BenchmarkBasicBatchRedis	1362045	1980634	🔴 -20x	🔴-14x
BenchmarkBasicBatchPostgres	86402994	308697749	🔵 2x	🔵10x
BenchmarkBasicTigerBeetleParallel	2946523	230546	-	-
BenchmarkBasicRedisParallel	38348	30556	🔴-76x	🔴-7x
BenchmarkBasicPostgresParallel	592111	1435170	🔴-4x	🔵6x

PASS
ok      kizzx2.com/tigerbeetle-benchmarks       204.673s

TL;DR

TigerBeetle is about 1 - 2 order of magnitudes slower than Redis / Postgres except for when batching 1000 transactions each batch, where it starts to catch up.

There is some speed up as suggested by running in parallel, but for the most part one would colloquially say that TigerBeetle is still drastically slower. Perhaps there are some highly specialized test case where this can differ which is not covered here.

Methodology

The first 2 scenarios are based on the samples code from the tigerbeetle-go client library.

The batch example was added based on the Basic example because it seemed to be losing too bad.

AMD benchmark was added because it was mentioned that TigerBeetle performs significantly worse on Mac than on Linux. I could confirm that on my particular Linux set up, TigerBeetle was just 100% slower than Postgres for non batch use case.

Parallel benchmark was added as suggested that it would theoretically lead to being able to witness the 1000x performance difference.

How to Run

1. Download TigerBeetle
2. Run the following between each benchmark run:

rm ./0_0.tigerbeetle
./tigerbeetle format --cluster=0 --replica=0 --replica-count=1 ./0_0.tigerbeetle
./tigerbeetle start --addresses=3000 ./0_0.tigerbeetle

3. Start Postgres and Redis

docker run --rm -it -p 5432:5432 -e POSTGRES_PASSWORD=postgres postgres
docker run --rm -it -p 6379:6379 redis

4. Run the benchmarks

go test -bench=. -count=10

5. 🍿