experiment-runner and EnergiBridge

This exercises will teach you how to use EnergiBridge together with the experiment-runner framework. Check the experiment-runner repo for an overview of the framework.

You will create an experiment using the framework and collect measures for different initial setups. The workflow for your experiment will be defined in the RunnerConfig.py file inside the experiment folder (see experiment-runner/examples for example experiments).

Setup

Start by creating your own RunnerConfig running (on your activated virtualenv):

$ python experiment-runner/ config-create examples/<your-example-name>

Exercise 1

For now we're going to stick with a simple experiment with one single factor and one single level. Later on we will see how to run more interesting experiments. The focus of this exercise is to teach you how to use EnergiBridge as a profiler within the experiment-runner framework. For this, we just want a single run of the following command:

$ sleep 5

We will measure the total CPU enery and the total memory used by this program.

Edit the create_run_table_model() method as follows:

def create_run_table_model(self) -> RunTableModel:
    """Create and return the run_table model here. A run_table is a List (rows) of tuples (columns),
    representing each run performed"""
    factor1 = FactorModel("example_factor1", ["example_treatment1"]) # just one factor with one level
    
    self.run_table_model = RunTableModel(
        factors=[factor1], data_columns=["total_cpu", "total_mem"] # we want to measure toatl cpu and memory
    )

    return self.run_table_model

Edit the stop_run() method as follows:

    def stop_run(self, context: RunnerContext) -> None:
        """Perform any activity here required for stopping the run.
        Activities after stopping the run should also be performed here."""
        
        output.console_log("Config.stop_run() called!")

        self.profiler.wait()

Draw inspiration from the examples folder to:

• Run the command $ sleep 3 using EnergiBridge at the start of the experiment (Hints: Edit start_run() and set self.profiler)
• Aggregate the raw data outputted by EnergiBridge to include it in the RunTableModel (Hints: edit populate_run_data() paying attention to the column names in the raw data)