This repository contains the scripts to produce the results of the energy budget paper. The full name of the paper and the co-athors are not given yet, as it is not published yet.
The raw results are not here, as they take several gigs. But you can reproduce them easily, following this guide.
The workflow is managed using Makefiles' rules. You need standard sciientific tool for the analysis of the results (R, ggplot2, python-pandas, python-scipy, python-rpy2).
We use the Batsim simulator . Clone it, compile it and install it. The algorithms used in the paper are within the simulator, as python scripts. Take a look at this file.
In the following of this file, we assume that this repository has been cloned in schedulers/pybatsim/schedulers/EXPE directory of batsim.
First generate the json files discribing each experiment by running:
mkdir expes
python generate.py
python generate_sdscblue.py
python generate_metacentrum.py
Then, you can start the simulation:
make simuls
Don't forget the -j option of make to run multiple simulations at once.
It tooks me 2 weeks to run everything on an old 256 cores machine (-j 256).
You can determine the progress of the simulations using:
echo "Simulations ended: " $(( $(ls expes/*/out_jobs* |wc -l) *100 / $(ls expes/ |wc -l))) "%"
Once the simulations are finished:
make results
This can also take quite a long time (about 1 day for me).
jupyter-notebook
Then, open TestnewMC.ipynb and rerun each nodes.