Part of the Q4 2022 instance of the Delft University of Technology Research Project, the concluding thesis for the Bachelor's of Computer Science and Engineering.
A modular simulator for testing and benchmarking caching algorithms, primarily for bipartite network algorithms. It was designed as part of a project to develop a bipartite version of the Online Mirror Descent caching policy. The title of the related paper is "Integral Caching using Online Mirror Descent in a Networked Context".
JSON files allow for configuring benchmark runs, from network topology over
catalog size and time horizon to the seed used for all randomness involved. As
a result, runs can be re-created perfectly with only the JSON file. All data
presented in the paper can be re-created using the config files provided in the
results folder.
In order to guarantee reproducibility, Pipenv is used. This package can be installed through Hombrew (macOS) or a variety of Linux package managers.
To run a config, first change the path set in run_from_config.py to point
towards the desired config. Then, run pipenv run python run_from_config.py.
Note that paths are relative from the directory
from which run_from_config.py is executed.