Use tests with realistic datasets/queries to quantitatively benchmark PIEDB. This repository contains the code of the PIEDB benchmark.
It includes:
multi-thread / multi-process mode (6 by default)
- 24-hour write stress test
- 24-hour query stress test, including random time window, random column tests
- write QPS: write 10 million+ data into database, when varying # of threads (1,2,3,8,16,32)
- read QPS: query 1-10 million+ data when varying # of threads
- mixed QPS: read while writing
- compare ratio of raw data to DB size, when varying compression algorithms
- create / remove database
- create / drop table
- table index
- create / drop views
- add / remove columns in tables
- insert
- update
- delete
- select from where
- table join
- limit
- groupby
- streaming clause
- time window: minute, hour, day
- math functions: count, sum, max, min, avg, etc