Production skew
A few customers make most orders and a few products take most pageviews. Hot keys and cold ranges are in the data.
Free datasets for benchmarks
Benchmark data with a skew and a clock.
Uniform random rows make every index look good. These tables have hot keys, weekday peaks and a long tail, and inserts that cluster in the working day and the busy season.
Database benchmark data with real distributions
Rows drawn uniformly at random spread the load evenly over every key and every hour. Every index looks selective and every cache looks warm. The classic benchmark schemas stop the clock.
Real distributions and live writes
Each dataset has the distributions its industry has. Concentration is a long tail, popularity drifts, and inserts cluster in the working day and the busy season. Load the download into Postgres, ClickHouse or SQLite. A download has the same bytes as the last one, and a rerun a month later measures the same input.
Get the benchmark data
Download the SQLite file or the CSV bundle and load it into the engine you are measuring. Reference columns have secondary indexes in the SQL targets. For larger volumes or a write load at your rate, leave an email above.
What you get
Time-correlated inserts
Inserts cluster in the working day and the busy season, so a time-ordered load replays the pattern a production write path sees.
Same bytes next run
Same dataset, same rows, a month later or on the other engine. The number that changed is the one you changed.
Volume on request
The published scale is one size. Tell us the row count you need and we generate it.
The datasets
Other use cases