Abstract

While machine learning models and algorithms are now often used for various network design, planning, provisioning, etc, not much 5G data is available. A Dynamic traffic generator is a valuable tool for evaluating and testing network performances.
In this thesis, we present a novel traffic generation framework that meets some of the most important 5G network requirements.
One of the distinctive features in 5G is the use of network virtualization, which allows network operators to partition the network into multiple "independent slices", each of which can carry different types of traffic, services, or applications such as enhanced Mobile Broadband, Ultra-Reliable Low Latency Communications, massive Machine Type Communications, VoIP, etc .
To meet the new 5G requirements a novel framework that addresses the key related problems is designed, i.e., modeling virtual network functions, allowing end-to-end measurement of the key performance indicators, with a realistic network traffic modeling. The results and evaluation show that our framework is a powerful traffic generator for the time being to test 5G networks.