The benchmark tool uses Mininet and the MPTCP kernel to test the congestion control algorithms implemented in the kernel.

These instructions should get you started on a Ubuntu server instance. Specifically we used Ubuntu server 18.04, MPTCP kernel v0.95, and Mininet v2.3.0d6. First install the MPTCP kernel on the machine, either using the apt repository or compiling the kernel yourself. More information about installing the MPTCP kernel can be fund here. Similar instructions about how to install Mininet can be found here.

Installing the kernel from the apt repository is the simplest method and only takes a couple of minutes.

sudo apt update
sudo apt-key adv --keyserver hkp://keys.gnupg.net --recv-keys 379CE192D401AB61
sudo sh -c "echo 'deb https://dl.bintray.com/cpaasch/deb stretch main' > /etc/apt/sources.list.d/mptcp.list"
sudo apt-get update
sudo apt-get install linux-mptcp -y

# after rebooting check if mptcp is installed
sudo dmsg | grep MPTCP

Compiling the kernel yourself allows you to change certain aspects of the kernel or specify that the MPTCP congestion control algorithms should be loaded per default and not as modules. Note that the compilation can take a long period of time, depending on the system. The -j parameter during make makes the compilation run in parallel and speed it up.

git clone --depth=1 git://github.com/multipath-tcp/mptcp.git
cd mptcp
make menuconfig
make -j $(nproc)
sudo make modules_install
sudo make install
sudo reboot

# after rebooting check if mptcp is installed
sudo dmsg | grep MPTCP

To install Mininet, simply clone the repository and use the convenience script Mininet provides.

git clone git://github.com/mininet/mininet
cd mininet
git checkout -b 2.3.0d6 2.3.0d6
cd ..
./mininet/util/install.sh -a

Depending on whether the jupyter analysis script also needs to run on the same machine, also jupyter and pandas needs to be implemented. Otherwise make sure at least to install iperf3, tcpdump, numpy.

sudo apt-get install -y iperf3 tcpdump python-pip htop
pip install numpy

After installing the MPTCP kernel and Mininet, the system should be able to run MPTCP experiments. Note that if the MPTCP congestion control algorithms are modules and not loaded, you need to use modprobe to make them available. Eg. sudo modprobe mptcp_coupled to load LIA.

After loading the necessary modules, the benchmark can be run with the following command:

sudo python main.py --topo two_paths --run de

For available options run python main.py -h.

Important options:

• topo: name of topology to use, points to JSON configs in folder topologies but can easily be adapted.
• run: which experiments to run
• cli: instead of running the experiment normally, run Mininet in CLI mode
• log: set the log level of Mininet
• no_dtcp: do not use tcpdump at all (no delay analysis possible)
• dtcp: keep packet trace file after experiment for further analysis