This is a fork of BenchBase, with the following enhancements:
Short decriptions of branches:
Minumum requirements for running the benchmark against YDB:
Above 1000 warehouses, the memory and CPU consumption grow linearly, you need:
E.g. to run 10000 warehouses you need to have at least 10 cores and 64 GB RAM. However, Instead of running 10000 warehouses on a single instance (and machine), we recommend to run at most 5000 warehouses per instance (preferably on separate machines).
To reduce memory consumption, make sure you don't use huge pages or transparent huge pages.
Prebuilt packages:
To build the benchmark you need Java >= 21. This project depends on the development version of ydb-jdbc-driver:
mvn clean install -DskipTests
Next, to build the benchbase-ydb package, execute the following command:
./mvnw clean package -P ydb -DskipTests
The simplest way is to use helper scripts from benchhelpers. You can find the full instruction here.
Here you can find description of the authentication. Usually you will either use anonymous authentication in case of self deployed YDB, or provide a service account key file using the saFile=file:
jdbc url parameter in tpcc_config_template.xml, when run managed YDB.
Please refer to this repository.