SaTE: Low-Latency Traffic Engineering for Satellite Networks

This repository accompanies our paper "SaTE: Low-Latency Traffic Engineering for Satellite Networks". It provides source codes, along with the necessary scripts, tools, and environment to reproduce the results discussed in the paper.

Overview

The repository includes:

• Source codes for SaTE the GNN system we proposed
• A pre-configured Singularity/Dokcer image that encapsulates all dependencies and tools needed for running the experiments.
• A collection of bash scripts located in satellite-te/utils/scripts/ that allow you to replicate the results presented in the paper.

Prerequisites

1. Hardware requirements

We conducted our experiments on a Standard_NC24ads_A100_v4 instance on Microsoft Azure and cluster at National Supercomputing Center Singapore (we requested 110gb memory and 1 A100 GPU for experiments).

2. Environment

To set up the environment, pull the image containing all necessary software dependencies:

singularity pull docker://hyizhak/sate_image_ot

Or create (and activate) the conda virtual environment with the environment.yaml we provide:

conda env create -f environment.yaml
conda activate satte

3. Runing Experiments

To run the experiments, please follow the steps below:

• Create a directory (e.g., .../raw/starlink).
• Download the Starlink input dataset from: https://drive.google.com/drive/folders/1h6kbOj4HpqofPNd7lkIJDTut4XF4ipAF?usp=sharing. Save the downloaded datasets (DataSetForSaTE25, DataSetForSaTE50, DataSetForSaTE75, and DataSetForSaTE100) into this directory.
• Update the variables INPUT_DIR and RAW_INPUT_DIR in utils/scripts/env to point to the directory containing the downloaded data.
• Run the corresponding adaptation scripts (e.g., utils/scripts/adapt_starlink.sh) to preprocess the dataset, including path pre-configuration and traffic aggregation.

Each experiment presented in the paper can be reproduced using the provided bash scripts. This repository consists mainly of:

satellite-te/
├── analyze/                        # Quick validation and result processing
├── baselines/                      # Baselines using Gurobi including LP, POP, TOP 
├── lib/                            # Source code of SaTE and another learning-based baseline Teal
├── output/                         # Trained models and testing reports
├── run/                            # Main classes scripts call
├── utils/
│   └── scripts/
│       ├── run_GS.sh               # Script for experiments on satellites linked with ground relays
│       ├── run_ISL.sh              # Script for experiments on satellites linked with laser links
│       ├── run_in_singularity.sh   # Script to facilitate experiments with singularity images
│       ├── log/                    # Logs from tasks on NSCC clusters             
│       └── ...
├── ...
└── README.md                       # This file

To execute an experiment, navigate to the directory and run the corresponding part in the bash file. For example to run experiments with ISL inter-satellite linkage mode:

cd satellite-te/utils/scripts
bash run_ISL.sh

We recommend to run with the container, for example:

bash run_in_singularity.sh run_ISL.sh