On the Robustness of Deep Learning-predicted Contention Models for Network Calculus

This repository contains the dataset used for the article "Graph-based Deep Learning for Fast and Tight Network Calculus Analyses" published in IEEE Transactions on Network Science and Engineering, and the paper "On the Robustness of Deep Learning-predicted Contention Models for Network Calculus" published at the 2020 IEEE Symposium on Computers and Communications and its [arXiv version]((https://arxiv.org/abs/1911.10522). We refer to the article and paper for a full explanation of the methodology used for generating the dataset.

Getting the dataset

The raw dataset can be accessed via the DOI: 10.14459/2019mp1524892. The following command can be used to download the full dataset via FTP:

$ wget -r ftp://m1524892:m1524892@dataserv.ub.tum.de/

Reading the dataset

The dataset is stored as serialized protobuf messages using pbzlib.

This repository contains an example python script for parsing the files. To get it and execute it:

$ git clone https://github.com/fabgeyer/dataset-deeptma-extension.git
$ cd dataset-deeptma-extension
$ pip3 install -r requirements.txt
$ python3 example.py path/to/dataset.train0.pbz

Alternative programming languages may be used with pbzlib (e.g. Java, Go).

Citation

If you use this dataset for your research, please include the following reference in any resulting publication:

@article{GeyerBondorf_TNSE2021,
	author        = {Geyer, Fabien and Bondorf, Steffen},
	journal       = {IEEE Transactions on Network Science and Engineering},
	title         = {Graph-Based Deep Learning for Fast and Tight Network Calculus Analyses},
	year          = {2021},
	volume        = {8},
	number        = {1},
	pages         = {75--88},
	doi           = {10.1109/TNSE.2020.3025806},
}

License

The data in this repository is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).