Elastic Slice-Aware Radio Resource Management with AI-Traffic Prediction

Network virtualisation and network slicing are the two essential innovations in the next generation of mobile networks also known as the 5G networks. Based on these innovations, multiple network slices with different requirements and objectives can share the same physical infrastructure. The techniques to efficiently allocate the available radio resources to different slices based on their requirements and their priority, also known as inter-slice radio resource management, has been the subject of many studies. The formerly proposed algorithms either assume the slices request maximum contracted data rates or they react passively as the demands arrive. This paper proposes to use Artificial Intelligence (AI) approaches to learn the pattern of the traffic demand of each network slices and predict the demands in the next decision interval. Based on the prediction of the slices' demands, a novel model for elastic inter-slice radio resource management is proposed to increase the multiplexing gain while not compromising the quality of offered connectivity services to the slices. The proposed model is evaluated using a practical scenario. The numeric results show that while the performance of the model under full demand is similar to former models, its elastic resource management enables more efficient resource allocation when the traffic demands vary over time.