Basic rdkafka_performance (from librdkafka/examples) benchmark.
In phase 1 it produces (no consumption during this phase) a fixed amount of data to a single partition topic, then in a second phase consumes it (no producton) with a single consumer.
Just run ./bench.
If you check the top of the script there are a few tunables you can override with environment variables.