Best Practices on Using ndnSIM¶
We highly recommend that ndnSIM users follow the principles below:
The latest version of the simulator should be used whenever possible.
The ndnSIM team is able to address users’s questions more efficiently when the latest released version of the simulator is used.
Users should fork and use the provided scenario template in order to implement their simulation scenarios.
The scenario template can be found here.
Examples of using the scenario template can be found here and here.
Scenario template structure
The scenario template contains the following directories:
- extensions: This directory is supposed to contain simulator extensions necessary for the written simulation scenarios. Specifically, these are any .cpp/.hpp files that will be linked to each of the included scenario.
- scenarios: Under this directory, the actual simulations are
supposed to be placed (each .cpp file becomes its own simulation
scenario – it must contain the
main
function).
How to compile a scenario template
To configure in optimized mode without logging (default):
./waf configure
To configure in optimized mode with scenario logging enabled (logging in NS-3 and ndnSIM modules will still be disabled, but you can see output from NS_LOG* calls from your scenarios and extensions):
./waf configure --logging
To configure in debug mode with all logging enabled
./waf configure --debug
If you have installed NS-3 in a non-standard location, you may need to set up
PKG_CONFIG_PATH
variable.How to run the simulation scenarios (without visualizer)
Normally, you can run scenarios either directly:
./build/<scenario_name>
or using waf:
./waf --run <scenario_name>
If NS-3 is installed in a non-standard location, on some platforms (e.g., Linux) you need to specify the
LD_LIBRARY_PATH
variable:LD_LIBRARY_PATH=/usr/local/lib ./build/<scenario_name>
or:
LD_LIBRARY_PATH=/usr/local/lib ./waf --run <scenario_name>
To run a scenario using the debugger, use the following command:
gdb --args ./build/<scenario_name>
How to run the simulation scenarios (with visualizer)
To run a scenario with the visualizer, you will have to setup some python environment variables to find the visualizer module. The easiest way to do so is by using the following commands:
cd ns-dev/ns-3 ./waf shell
After these commands, you will have the complete environment to run the vizualizer.
The following command will run a scenario with the visualizer:
./waf --run <scenario_name> --vis
or:
PKG_LIBRARY_PATH=/usr/local/lib ./waf --run <scenario_name> --vis
If you need automatic node placement, set up an additional environment variable:
NS_VIS_ASSIGN=1 ./waf --run <scenario_name> --vis
or:
PKG_LIBRARY_PATH=/usr/local/lib NS_VIS_ASSIGN=1 ./waf --run <scenario_name> --vis
Users should modify the README file of their fork of the scenario template repo to clearly note the version/fork of ndnSIM that their scenarios are supposed to work in.
Our effort is to keep ndnSIM updated with the latest advancements of NDN research. As a result, the ndnSIM API might change from one release to another, while the API of a new release might not be fully compatible with the API of previous releases. Therefore, we strongly encourage you to note the specific version (commit hash of the ndnSIM repository and base NS-3 repository) that was used for your simulations. This will guarantee the reproducibility of your results, which is critical for simulations that are published in scientific papers.
Note
We highly encourage that users share their simulation scenarios with the rest of the ndnSIM users’ community. To achieve that, please let us know if you would like us to fork your simulation scenarios under the named-data-ndnSIM GitHub organization.