Benchmarks

Published. Reproducible. Independent.

We publish every benchmark we run — methodology, hardware, commands, and raw timing data. If a number isn't reproducible, it isn't a benchmark.

Most database tools publish a single headline number on the landing page and never show their work. We do the opposite. Every benchmark we cite — every 89×, every 8-minute migration, every sustained throughput claim — has a full write-up below with the hardware, the workload, the exact commands we ran, and the raw timing.

If you can reproduce it, it's real. If you can't, it shouldn't be on a marketing page.

The headline numbers.

Each one is linked to a full write-up below.

89×

faster Postgres → Snowflake (1B rows)

188×

faster Postgres → Firebolt (1B rows)

8m 56s

MySQL → Snowflake, 1 billion rows

53s

MySQL → Firebolt, 10 million CDC ops

Every benchmark we've published.

14 write-ups across Wirekite Data and Wirekite Replicate. Click any card for the full methodology, hardware, and raw timing.

Data

MySQL → Firebolt

1B rows in 9m 53s

Read benchmark

Data

MySQL → Snowflake

1B rows in 8m 56s

Read benchmark

Data

Oracle → AlloyDB

1B rows

Read benchmark

Data

PostgreSQL → Firebolt

188× faster, 1B rows

Read benchmark

Data

PostgreSQL → Snowflake

89× faster, 1B rows

Read benchmark

Data

SQL Server → Firebolt

1B rows in 12m 19s

Read benchmark

Data

SQL Server → Snowflake

1B rows in 10m 52s

Read benchmark

Replicate

MySQL → Firebolt

10M CDC ops in 53s

Read benchmark

Replicate

MySQL → Snowflake

10M CDC ops in 2m 41s

Read benchmark

Replicate

Oracle → Firebolt

10M CDC ops in 5m 50s

Read benchmark

Replicate

Oracle → Snowflake

10M CDC ops in 6m 7s

Read benchmark

Replicate

PostgreSQL → Firebolt

10M CDC ops in 2m 18s

Read benchmark

Replicate

PostgreSQL → Snowflake

10M CDC ops in 3m 52s

Read benchmark

Replicate

SQL Server → Firebolt

10M CDC ops in 1m 13s

Read benchmark

How we benchmark.

01 Same hardware on both sides. Source and target run on identical instance types. The benchmark measures Wirekite's pipeline, not network speed differentials.
02 Two-server topology. Source database on one server, target on another, Wirekite running on either or both. Mirrors a real production deployment.
03 Default product settings. No tuning that isn't in the docs. If a customer downloads the product and runs it, they get the published number.
04 All commands published. Each write-up includes the schema, the workload generator, the exact Wirekite command lines, and the queries used to verify the result.

Want a benchmark for your specific source × target?

We add benchmarks based on what teams ask for.

Request a benchmark