🚦 TrafPy Packing Algorithm Speedup 🐎

Implementation of the vectorised traffic flow packing algorithm reported in A Vectorised Packing Algorithm for Efficient Generation of Custom Traffic Matrices.

Note: This algorithm has already been integrated into the full TrafPy open source tool as the default packer for traffic generation - refer to the main TrafPy GitHub repository for documentation and further details on how to use TrafPy traffic generation for your own projects.

Setup

Open your command line. Change the current working directory to the location where you want to clone this project, and run:

$ git clone https://github.com/cwfparsonson/trafpy_vectorised_packer

In the project's root directory, run:

$ python setup.py install

Then, still in the root directory, install the required packages with conda (env name defined at top of .yaml file):

$ conda env create -f requirements/default.txt

Running a Quick Packing Speed Experiment

First, set the hyperparameters you want to use in scripts/configs/traffic_generation_default.yaml. For example, to use the original packer, set flow_packer_cls: trafpy.generator.src.packers.flow_packer_v1.FlowPackerV1. To use the vectorised packer, set flow_packer_cls: trafpy_vectorised_packer.vectorised_flow_packer.VectorisedFlowPacker, which will run the new vectorised packer originally written in this repository (N.B. this has now been integrated as the default packer in the TrafPy library as trafpy.generator.src.packers.flow_packer_v2.FlowPackerV2).

Next, from the root directory of where you cloned this repository, run:

$ python scripts/packer_speed_test.py

Running a Hyperparameter Sweep

This repository makes use of Weights and Biases for experiment tracking and hyperparameter tuning.

To test the different packers over different traffic generation conditions and network sizes, first set the hyperparameters you want to use in scripts/wandb_sweep_config.yaml.

Then, start a new tmux session called trafpy_sweep:

$ tmux new -s trafpy_sweep

Next, inside the trafpy_sweep tmux session and from the scripts/ directory of this cloned repository, run the sweep script with n = 2 parallel sessions and a d = 30 second delay between launching each of these sessions to avoid read-write conflicts:

$ python run_wandb_sweep.py -s trafpy_sweep -n 2 -d 30

Citing this work

If you find this project or the associated paper useful, please cite our work:

@article{parsonson2023vectorised,
    title = {A Vectorised Packing Algorithm for Efficient Generation of Custom Traffic Matrices},
    journal = {OFC'23: Optical Fiber Communications Conference and Exhibition},
    year = {2023},
    url = {https://arxiv.org/abs/2302.09970},
    author = {Christopher W.F. Parsonson and Joshua L. Benjamin and Georgios Zervas},
}