Running pbench¶
At this point, you should already have a Presto C++ or Presto Java cluster running.
1. Download pbench¶
1.1 Download the pbench tar¶
Download the pbench
tar for your platform below. This contains the compiled pbench
binary and the relevant benchmark
configuration files.
1.2 Extract pbench¶
Extract the downloaded tar.gz file and change into the created pbench
directory in a
new terminal window.
cd pbench
2. Run pbench¶
Once in the pbench
directory, run pbench with ./pbench run
and specify the benchmark configuration files
related to the run. This example uses the sf1
scale factor and ds_power
run flavor, which runs all 99 TPC-DS queries
sequentially. Change the catalog
to hive
and schema
to tpcds
in the json
configuration file benchmarks/tpc-ds/sf1.json
.
For Presto C++:
./pbench run benchmarks/native_oss.json benchmarks/tpc-ds/sf1.json benchmarks/tpc-ds/ds_power.json
For Presto Java:
./pbench run benchmarks/java_oss.json benchmarks/tpc-ds/sf1.json benchmarks/tpc-ds/ds_power.json
You should see logs for each query being submitted and the results, including execution time and row count.
Supplying different json files allow you to run different benchmarks. For more information on this format, visit the pbench wiki.
Troubleshooting¶
If you see a permissions pop-up that prevents running pbench
on MacOS, run the following command on the downloaded
.tar.gz
file before extracting:
xattr -d com.apple.quarantine pbench_darwin_arm64.tar.gz
pbench
file you downloaded.