R2lab Notebooks
Jupyter notebooks for reproducible research
We are convinced that the general approach of writing Jupyter notebooks is an effective means to improve research reproducibility.
In this section we give some links to runnable notebooks, with the hope to foster this medium as a common practice.
Hosting at mybinder
A notebook requires a so-called "kernel", typically a python interpreter, and thus computing resources.
There are several options for running such a document :
install jupyter locally and run the notebook from your computer; this also implies that you install all the dependencies for that experiment,
or run the notebook from a dedicated infrastructure;
mybinder.orgis one such infrastructure that lets us easily expose such dedicated notebook services from just a github repository.
Hosted notebooks, ssh keys and authentication
These two approaches have their obvious pros and cons.
The hosted approach allows for instant gratification, but on the other hand has more limitations, and in the case of R2lab in particular, it will not let you actually log into the testbed, because the docker container that runs behind your mybinder notebook has no knowledge of your ssh credentials.
So going for a local installation is a little more tedious, but offers more capabilities.
Building a radiomap
Our first example of a runnable notebook does not address a research theme exactly, but hopes to serve as a methodological guidelines on a representative subject. Our pretext here is to gather calibration-oriented data, i.e. perceived power at node (a) when node (b) is emitting, for all couples of nodes.
You can find the source for this notebook:
under github in the r2lab-demos area. This repo contains all the needed code to collect your own data; because of the limitations exposed above, the repository also contains some gathered data, so that visualization can be done right away from mybinder as well
in mybinder here : radiomap
Comparing batman vs OLSR
Another example, similar to radiomap, was about gathering data in order to comparatively assess these 2 mesh routing protocols. It can be found here:
You can find the source for this notebook:
under github in the r2lab-demos area. This repo contains all the needed code to collect your own data; because of the limitations exposed above, the repository also contains some gathered data, so that visualization can be done right away from mybinder as well
in mybinder here : radiomap