Running Benchmarks with Data Collection Tools

[safl]

When running benchmarks, it can be useful to collect data in parallel using auxiliary tools. Some tools can be prefixed to the benchmark command, while others may require separate start and stop commands. This section provides an example of how to use a tool in parallel with a benchmark run, storing collected data in a log file.

Example: Running a Data Collection Tool in Parallel

In this example, we use a data collection tool launched via screen, enabling it to run in a separate session while the benchmark tool executes. This setup allows you to start data collection before the benchmark begins and stop it once the benchmark completes.

1. Start the data collection tool:

   Before running the benchmark, start the tool in a detached screen session:

screen -dmS collector data_collection_tool >> /tmp/collector.log

• screen -dmS collector: Starts a new detached screen session named "collector."
• data_collection_tool >> /tmp/collector.log: Runs the data collection tool, appending output to /tmp/collector.log.

2. Run the benchmark tool:

   Execute your benchmark tool as desired. The data collection tool will run concurrently in the background.

3. Stop the data collection tool:

   After the benchmark finishes, stop the screen session running the data collection tool:

screen -S collector -X quit

This command terminates the "collector" session, stopping the data collection tool.

4. Retrieve the collected data:

   Use cijoe to fetch the log file with the collected data:

cijoe.get("/tmp/collector.log")

This approach allows for flexible data collection without interrupting the benchmark process, and it makes it easy to access the collected data for further analysis.