Simulation server cluster (discrete-event simulation). Chosing an optimal C value (number of specialized clusters) depending on arrival rate.

How to use

usage:

main.py [-h] [--simulation_time SIMULATION_TIME] [--num_runs NUM_RUNS] [--min_runs MIN_RUNS] [--confidence_level CONFIDENCE_LEVEL] [--best_c | --no-best_c]

options:

-h, --help            show help message and exit
--simulation_time SIMULATION_TIME runtime of each individual simulation
--num_runs NUM_RUNS   number of simulations to run with a fixed set of parameters
--min_runs MIN_RUNS   minimum number of successful runs needed to calculate statistics
--confidence_level CONFIDENCE_LEVEL confidence level
--best_c  only run simulations for lambda=1 and determine the best C

default values:

Simulation time: 500
Number of runs: 10
Minimum number of runs: 5
confidence level: 95%
best C (running simulation for lamba=1 only): False

structure

simulation.py contains the code for the simulations.
main.py contains a minimal CLI argument parser.

A class is used to represent a simulation, 2 wrappers are then used. One to run the simulations and produce statistics related to the mean response time, one to produce statistics related to the mean loss rate.
Multiprocessing library is used for parallelism, as multiple simulations are needed with each set of parameters they can be executed in parallel.