This repository contains the artifact accompanying the following paper:

Romain Jacob, Jackie Lim, and Laurent Vanbever. 2023.
Does rate adaptation at daily timescales make sense?.
In 2nd Workshop on Sustainable Computer Systems (HotCarbon ’23),
July 9, 2023, Boston, MA, USA. ACM, New York, NY, USA, 7 pages.
https://doi.org/10.1145/3604930.3605713

We provide a conda environment file to install all depencies required.

# Re-create the required environment
conda env create -n OVH --file environment.yml
# Activate it
conda activate OVH

The hotcarbon23.ipynb walks through the analysis presented in the paper. Most likely, that is where you should start.

All plots are saved within the notebook, to allow visualization without re-running it. The downside is that the file is fairly large (50MB) and GitHub won't let you view it. You can use nbviewer instead.

All the scripts used for the analysis are provided:

• helpers/download_OVH.py allows to conveniently download the entire OVH dataset (beware, it's about 54GB).
• parse_*.py are the scipts used to parse the dataset (not all are used at the moment). These scripts are not optimized; re-running them on the entire dataset takes a couple of days and a lot of memory. (Re)use at your own risk.
• *.csv are the script outputs, which we provide for convenience.