A Simulation Execution Manager for ns-3

This is a Python library to perform multiple ns-3 script executions, manage the results and collect them in processing-friendly data structures. For complete step-by-step usage and installation instructions, check out readthedocs.

How to cite us

If you used SEM for your ns-3 analysis, please cite the following paper, both to provide a reference and help others find out about this tool:

Davide Magrin, Dizhi Zhou, and Michele Zorzi. 2019. A Simulation Execution Manager for ns-3: Encouraging reproducibility and simplifying statistical analysis of ns-3 simulations. In Proceedings of the 22nd International ACM Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWIM '19). ACM, New York, NY, USA, 121-125. DOI: https://doi.org/10.1145/3345768.3355942

Contributing

If you want to contribute to sem development, first of all you'll need an installation that allows you to modify the code, immediately see the results and run tests.

Building the module from scratch

This module is developed using pipenv: in order to correctly manage virtual environments and install dependencies, make sure it is installed. Typically, the following is enough:

pip install -U pipenv

Note that, depending on the specifics of your python installation, you may need to add ~/.local/bin to your path. In case this is needed, pip should warn you during installation.

Then, clone the repo (or your fork, by changing the url in the following command), also getting the ns-3 installations that are used for running examples and tests:

git clone https://github.com/DvdMgr/sem
cd sem
git submodule update --init --recursive

From the project root, you can then install the package and the requirements with the following:

pipenv install --dev

This will also get you a set of tools such as sphinx, pygments and pytest that handle documentation and tests.

Finally, you can spawn a sub-shell using the new virtual environment by calling:

pipenv shell

Now, you can start a python REPL to use the library interactively, issue the bash sem program, run tests and compile the documentation of your local copy of sem.

Running tests

This project uses the pytest framework for running tests. Tests can be run, from the project root, using:

python -m pytest --doctest-glob='*.rst' docs/
python -m pytest -x -n 3 --doctest-modules --cov-report term --cov=sem/ ./tests

These two commands will run, respectively, all code contained in the docs/ folder and all tests, also measuring coverage and outputting it to the terminal.

Since we are mainly testing integration with ns-3, tests require frequent copying and pasting of folders, ns-3 compilations and simulation running. Furthermore, documentation tests run all the examples in the documentation to make sure the output is as expected. Because of this, full tests are far from instantaneous. Single test files can be targeted, to achieve faster execution times, by substituting ./tests in the second command with the path to the test file that needs to be run.

Building the documentation

Documentation can be built locally using the makefile's docs target:

make docs

The documentation of the current version of the package is also available on readthedocs.

Running examples

The scripts in examples/ can be directly run:

python examples/wifi_plotting_xarray.py
python examples/lorawan_parsing_xarray.py

Troubleshooting

In case there are problems with the pandas installation (this will happen in macOS, for which no binaries are provided), use the following command for installation (and see this pandas issue as a reference):

PIP_NO_BUILD_ISOLATION=false pipenv install

Authors

Davide Magrin