Networking Dataset

Dataset Overview

Background

• The dataset simulates network scenarios, including three aspects: network topology, routing configuration, and source‑destination traffic matrices.
• The focus is on predicting the average per‑packet delay for each source‑destination path.
• The training set comprises networks with 25–50 nodes, the validation set includes networks with 51–300 nodes, and the test set is of comparable size to the validation set.

Dataset Contents

routings Directory

• Stores routing configuration files; each record contains a list of nodes representing a source‑destination path.

graphs Directory

• Contains network topology files described in Graph Modeling Language (GML) that specify nodes and links.

input_files.txt

• Lists the topology and routing files for each simulation.

traffic.txt

• Provides traffic parameters used in the simulations, including maximum average Lambda and path parameters.

simulationResults.txt

• Records simulation outcomes for each sample, including global statistics and path‑specific metrics.

stability.txt

• Supplies additional information on dataset stability, primarily the simulation time required to reach stable conditions.

linkUsage.txt

• Reports link‑usage statistics for each source‑destination pair, covering port utilization and loss rates.

Data Processing

Preprocessing

• The preprocessing.py script processes the raw data files, extracting and organizing traffic measurements, simulation results, and link‑usage information.

Neural Network Data Handling

• Additional processing adapts the data for spiking neural networks, storing the prepared data in AWS S3 and executing processing scripts on SageMaker.

Neural Network Execution

• Models such as snn_leaky.py and snn_synaptic.py are used to construct spiking neural network architectures for network performance prediction.