AQM (qdisc) Evaluation using flent tests

Requirements Setup

nitk-nest

python3 -m pip install nitk-nest

flent

sudo add-apt-repository ppa:tohojo/flent
sudo apt install flent

Install iproute2 (version >= 5.5) and apply the custom flent patches as mentioned here
Clone/Download the repository

git clone https://github.com/shashank68/aqm-evaluation

Usage

Use the --help option to view the configurable network parameters

sudo python dumbbell_flent.py --help

usage: dumbbell_flent.py [-h] [--rtt RTT] [--ecn {Yes,No}] [--duration TIME] [--bottleneck_bw BANDWIDTH] [--qdisc {fq_codel,fq_pie,codel,pie,cake}] [--no_offloads {Yes,No}] [--number_of_tcp_flows NUM] [--qdelay_target TARGET] [--results_dir DIR]

optional arguments:
  -h, --help            show this help message and exit
  --rtt RTT             Round trip time for flows (ms)
  --ecn {Yes, No}        Turn on ECN
  --duration TIME       Duration of test in seconds
  --bottleneck_bw BANDWIDTH
                        Bottleneck bandwidth (mbit)
  --qdisc {fq_codel, fq_pie, codel, pie, cake}
                        AQM algorithm (qdisc)
  --no_offloads {Yes, No}
                        Turn off GSO, GRO offloads
  --number_of_tcp_flows NUM
                        Number of flows in tcp_nup flent test
  --qdelay_target TARGET
                        Queue delay target (For qdisc)
  --results_dir DIR     Directory to store the results

The default values for the above parameters are defined in exp_config.py.

Example usage to run a test using the specified parameter values:

sudo python3 dumbbell_flent.py --rtt=100 --bottleneck_bw=80 --qdisc=fq_codel --ecn=No --no_offloads=Yes --results_dir=foo/bar

The above script will create a result directory named using the parameter values specified.

AQM_FLOWS_BANDWIDTH_RTT....DATE-TIME

Running multiple tests with different combinations of the parameter values

To run tests with custom combination of parameters, create a combinations_config.json from the example given

cp combinations_config.json.example combinations_config.json

Modify the values in combinations_config.json if required
To execute tests with the configured parameters run:

sudo python run_all_combinations.py

If you do not create a combinations_config.json file then the default values in run_all_combinations.py will be used