Benchmark Runner

Test the performance of joining a bunch of tables

Build

make build

Run

The database to target is specified by the standard postgres environment variables. You'll also need to either set PGPASSWORD or configure a .pgpass file

PGHOST="localhost" PGDATABASE="join_test" PGUSER="brian" PGPASSWORD="pass" ./run_with_docker.sh input.json

Input

Join-bench takes a json file that describes the benchmarks to run.

$ cat input.json
[
    {
        "join-type": "chained",
        "max-tables": 10,  # Queries will start by joining 2 tables, increasing by one until all tables are joined.  Number of tables joined will be the X axis on the plot.
        "max-rows": 10000,  # Benchmarks will be performed at 10 rows, 100 rows, etc. until max-rows is reached.  Creating a separate line on the plot for each.
        "extra_columns": 2,
        "max_id": 5,
        "create-indexes": true,
        "output-filename": "benchmark_1",
        "plot-title": "My Chained Benchmark Title"
    },
    {
        "join-type": "enums",
        "max-rows": 10000,  # Benchmarks will be performed at 10 rows in the primary table, increasing by a factor of 10 until max-rows is reached
        "max-enums": 100,  # Queries will start by selecting (and optionally filtering by) 1 enum column, increasing by one until max-enums is reached
        "possible-enum-values": 10,
        "extra-columns": 2,
        "where-clause": true,
        "output-filename": "benchmark_1",
        "plot-title": "My Enum Benchmark Title"
    },
    {
        "join-type": "foreign-keys",
        "max-primary-table-rows": 10000,  # Benchmarks will be performed at 10 rows in the primary table, increasing by a factor of 10 until max-rows is reached
        "max-fk-tables": 100,  # Queries will start by selecting from (and optionally filtering by) 1 foreign key table, increasing by one until max-fk-tables is reached
        "fk-rows": 100,
        "fk-extra-columns": 2,
        "extra-columns": 2,
        "where-clause": true,
        "output-filename": "benchmark_1",
        "plot-title": "My Foreign Key Benchmark Title"
    }
]

Output

Benchmark plots and CSV files are stored in the results/ directory by default