Field kit

The rankings get the attention. This page is what we run to produce them: the loadgen, the price math, the spreadsheets that never make it into the article. Nothing here pays us, and anything we stop reaching for gets dropped from the page.

for the benchmarks

hyperfine. Command-line benchmarking with warmups and proper statistics. Half the numbers on this site started as a hyperfine run.

k6. Load tests as scripts you can read a year later. When a ranking says "at 500 requests per second", this is what was doing the requesting.

pgbench. The boring standard for database throughput. Boring is the point; boring is comparable.

for the price math

DuckDB. Every pricing teardown is a pile of CSV exports and one DuckDB session. It reads anything and never argues.

Visidata. Terminal spreadsheet for the first pass over a vendor's pricing export, before the numbers deserve a real script.

for keeping us honest

git. Criteria files are committed before scoring starts. The history is the receipt.

Wayback Machine. Pricing pages change quietly. Archive first, cite the snapshot, sleep fine.