Demonstration of Real-Time Traffic Forecast on a Live 5G Testbed

Network Virtualization is considered as a main pillar in the architecture of future wireless communication networks. Virtualization is fundamental for the implementation of different applications and key technologies of 5G and future networks like Network Slicing, Edge Computing, and Software-Defined Networking (SDN). In such complex systems, network automation is essential and can be leveraged to overcome network management challenges. Network traffic prediction is one of the most important proactive approaches in network management and planning. This work proposes a real-time Testbed for wireless network traffic forecasting based on pre-trained Deep Learning (DL) models. A virtualized 5G core network is examined as an environment for both training and test purposes. The prediction of the network traffic is performed while the User Equipment (UE) is performing various activities. In addition to the results of the forecasting models, suggestions for future improvements of the Testbed and possible utilization fields are discussed.