kesselrun

A discrete-event simulation framework for modeling serverless workflows across the 3D Compute Continuum. Allows the modeling of workflows and their tasks by configuring execution time and bandwidth among others. Each layer of the 3D Compute Continuum can be configured independently.

• \framework contains the framework
• \simulations contains concrete usages (simulations) of the framework. nsga.py is the main simulation with multi-objective optimization using NSGA-II.
  • GoogleProofSimulation.py is a simulation that models a Google Cloud serverless workflow and is used for validation
  • RequestPatternSimulationTest.py is a simulation to show different request patterns and how to implement custom patterns
  • BayesianSimulation.py is the predecessor of nsga.py and uses bayesian optimization. It is kept for reference and was not used in the final thesis.
  • Simulation.py is a simple simulation without the optimization, used for running single simulations.

---

Usage

Look into main.py and Simulation.py for a simple example of how to use the framework.

To run the NSGA-II multi-objective optimization simulation run nsga.py. The amount of simulations can be controlled by changing the following lines:

algorithm = NSGA2(pop_size=90)  # adjust population size
termination = get_termination("n_gen", 111)  # adjust generations

which results in 90 * 111 = 9990 simulations.

I recommend pymoos documentation https://pymoo.org/getting_started/index.html