5. Implementation

5.1 Software or programming language

The simulation model was developed using python 3.8 and simpy 4.0. Simpy details are here: https://simpy.readthedocs.io/en/latest/

The exact software versions are:

joblib=0.15.1
jupyterlab=3.0.9
matplotlib=3.3.4
numpy=1.19.2
pandas=1.2.3
pip=21.0.1
python=3.8.12
scipy=1.6.1
simpy=4.0.1

A conda virtual environment is provided to manage versions on a local machine.

5.2 Random sampling

All sampling uses numpy.random.Generator. A numpy generator object implements the Permuted Congruential Generator 64-bit (PCG64; period = \(2^{128}\); maximum number of streams = \(2^{63}\)).

Repeatable experiments and common random number streams are used in the model. Non overlapping streams are creating using a numpy.random.SeedSequence. One seed - the replication number - is passed to a SeedSequence and \(n\) child seeds are spawned that are then used to create the \(n\) streams. The seeds spawned by SeedSequence ensure that these streams are “very likely” non-overlapping.

5.3 Model execution

Simpy implements a process based simulation worldview.

5.4 System Specification

The model was coded and run on Intel i9-9900K CPU with 64GB RAM running the Pop!_OS 20.04 Linux.