5G E2E Network Slicing Predictable Traffic Generator

Automated resource management for 5G network slicing implies the need to assign each slice the necessary resources, i.e., the ability to predict their respective requests and resource requirements. Machine learning models and algorithms can meet these needs provided the required data is available. Unfortunately, 5G traffic data remains sparse despite many studies relying on machine learning models and algorithms for traffic forecasting or automated network resource management. In this study, we introduce a 5G-type predictable traffic generator that relies on the refactoring of open data of vehicle and pedestrian traffic from the City of Montreal. Indeed, the latter data is refactored in order to generate different classes of network traffic, with different characteristics associated with typical 5G applications, and then with different traffic patterns and peak hours. The result is a valuable traffic generation tool for researchers interested in validating machine learning algorithms aimed at, for example, traffic forecasting, resource elasticity, or automated scaling of slice resources.