This repository contains part of the code for the paper "DeepMPLS: Fast Analysis of MPLS Configurations Using Deep Learning" published at the IFIP Networking 2019 conference. For access to the dataset only, please refer to the dataset repository. The tools currently only supports the query types specified in the DeepMPLS paper.

The dataset is stored in the dataset repository using git lfs. Install git lfs on your system first and then clone the code and dataset repository using:

$ git lfs clone --recursive https://github.com/fabgeyer/deepmpls.git
$ cd deepmpls

To install the required python dependencies, use:

$ pip3 install -r requirements.txt

The repository contains an implementation of the Graph Neural Network used in the paper based on PyTorch Geometric. Currently the neural network can only be used to predict the satisfiability of a query.

Usage for training on the paper's dataset:

$ python3 neural_network_qpred.py

In order to only partially load the dataset, the nnetworks argument can be used to specify the number of networks to load:

$ python3 neural_network_qpred.py --nnetworks 10

The repository contains also a simple command line utility for transforming MPLS networks to their DeepMPLS graph representation. It uses the XML file format used by P-Rex for representing the topology and the MPLS configuration.

Usage:

$ python3 graph_transformation.py <topo.xml> <routing.xml> '<a> b <c>' k

Example:

$ python3 graph_transformation.py P-Rex/test/test_cli/1/topo.xml P-Rex/test/test_cli/1/routing.xml '<.*> s1 .* s7 <>' 2

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

@inproceedings{GeyerSchmid_Networking2019,
	author    = {Geyer, Fabien and Schmid, Stefan},
	title     = {{DeepMPLS}: Fast Analysis of {MPLS} Configurations Using Deep Learning},
	booktitle = {Proceedings of the 18th IFIP Networking Conference},
	year      = {2019},
	month     = mai,
	address   = {Warsaw, Poland},
	doi       = {10.23919/IFIPNetworking.2019.8816842},
}