This repository is in a working in progress state.
If you are looking for the source code of the experiments of Emanuele
Petriglia's master's thesis, discussed in October 2024, see the
petriglia-thesis-2024
branch.
The thesis, a summary and the presentation slides are available in another repository hosted on GitLab, but they are written in Italian.
configs
: contains the specific configuration for the environment, models and other aspects of the experiments.dataset
: contains the dataset used to have real function invocations traces in the DFaaS environment (instead of generated ones).notebooks
: Python Jupyter notebooks used mainly to show plots of the experiments and do some simple prototyping or experiments.patches
: required custom patches to Ray RLlib or other libraries needed to run the experiments.plots
: non-interactive Python scripts to generate plots from experiments.results
: default directory where the experiment data is stored. This folder is not shown in the repository because it contains ephemeral data.tests
: some Python scripts used to test the Ray RLLib, the DFaaS environment, or other miscellaneous tests.
The experiments are run and tested on Ubuntu 24.04 using Python 3.12. For a
reproducible development environment, it is preferable to install the
dependencies in a virtual environment (see the
venv
module). The venv
module is not installed by default in Ubuntu, it must be installed using sudo apt install python3.12-venv
.
To run the experiments, the dependencies are:
-
Ray RLlib: this is a reinforcement learning library used to define the DFaaS custom environment, run the experiments by training the models with the implemented algorithms. The version is pinned to 2.40.0.
-
PyTorch: is a library for deep learning on GPUs and CPUs. It is used by Ray RLlib when training models with deep learning reinforcement learning algorithms.
-
orjson: is a JSON library that is faster than the standard library. Used because the experiments generate large JSON files that slow down the encoding/decoding processes.
Only these dependencies need to be installed manually. pip
automatically
installs the dependencies, and some of them are directly used by the project
(like NumPy, Gymnasium, Pandas or NetworkX).
The following dependencies are required for plotting or running notebooks:
-
Matplotlib: is a plot generation library used in the scripts in the
plots
andnotebooks
directory. -
Jupyter Notebook: Jupyter notebooks are used to create plots and explore results and statistics from the experiments. The notebooks are stored in the
notebooks
directory. -
ipympl: this is an extension for Jupyter Notebook to support interactive Matplotlib using Jupyter Widgets
ipywidgets
. -
nbstripout: utility that erases a Jupyter Notebook's output before committing to git.
Note that scripts in the plots
directory are not interactive and only require
Matplotlib.
There are two requirements files for pip
in the repository:
requirements.base.txt
: contains only the dependencies listed above with fixed versions,requirements.txt
: contains the full list of dependencies with fixed versions.
Run the following commands to set up the development environment with Ubuntu:
$ sudo apt install python3.12-venv
$ git clone https://github.com/unimib-datAI/marl-dfaas.git
$ cd marl-dfaas
$ python3.12 -m venv .env
$ source .env/bin/activate
$ pip install --requirement requirements-base.txt
Or, for perfect reproducibility:
$ pip install --requirement requirements.txt
Please note that the requirements file expects a machine with an NVIDIA GPU and CUDA (at least 12.4) installed for PyTorch. PyTorch can also be used with a CPU, in this case follow the instructions on the official website.
The requirements.txt
also contains some development tools:
black
: a source code formatted for Python code and Jupyter notebooks,pylint
: a static code analyser,pre-commit
: to run hooks when doing a Git commit,nbstripout
: to clean the output of notebooks before committing them.
WIP
Important: always run Python scripts from the project root directory to
allow loading of commonly used modules (dfaas_env.py
...). As example, if you
need to run a test script:
$ python tests/env/local_strategy.py
Just run:
$ jupyter notebook --config jupyter_notebook_config.py
Then open http://localhost:8888/tree in a Web browser.
The selected version of Ray RLlib needs to be patched to fix some bugs or
undesirable behaviour that has not yet been addressed upstream. The patches are
collected in the patches
directory and can be applied using the
patch
command:
patch -p0 < patches/NAME.patch
The patches have only been tested with Ray 2.40.0. They will only work if the
virtual environment is named .env
and the Python version is 3.12, as the file
path is hardcoded into the patch file.
Note: The patch
binary is required and preinstalled on Ubuntu. If not
available, it can be installed with apt install patch
.
The patches are created using the standard
diff
tool:
diff -Naru .env/.../rllib/example.py .env/.../rllib/example_new.py > patches/NAME.patch
See this reply on StackExchange for more information.
Copyright © 2024-2025 Emanuele Petriglia
The source code in this repository is licensed under the Apache License, version 2.0. See the LICENSE file for more information.
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.