This section includes results of standard benchmark tests TPC-H, TPC-DS, and ClickBench for pgpro_metastore.
Test details:
Data size: 100 GB. It corresponds to the scale factor of 100 in the benchmark test configuration.
Server: 64 vCPU cores and 128 GB of RAM.
Limitations:
Postgres Pro configuration parameters did not impact benchmark test results.
The duckdb.max_memory
configuration parameter was set to 32Gb.
The duckdb.threads
configuration parameter was set to 8.
In addition to the query execution time, the system metrics described in the table below were measured.
|
Metric |
Units of Measurement |
Description |
|---|---|---|
|
|
Seconds |
The number of seconds the CPU was used by a process running in user mode |
|
|
Seconds |
The number of seconds the CPU was used by a process running in core mode |
|
|
MB |
The maximum amount of memory occupied over process execution time |
|
|
MB |
The average amount of memory occupied over process execution time |
|
|
MB |
The amount of data read from a storage over process execution time |
|
|
MB |
The amount of data written to a storage over process execution time |
Preparation of the test data and tables:
The test data in the CSV format was generated.
For more information about used instruments and the overall procedure, refer to https://tpc.org.
CSV files were uploaded to heap tables.
Test data was exported to Parquet files, then analytical views were created for them.
When creating analytical views, column data types corresponding to column data types of Postgres Pro tables were specified. It is required for the correct functioning of analytical queries.