5G Services Placement in Dynamic Multi-clusters

The use of Machine Learning (ML) tools in 5G networks optimization problems is challenged by dynamical changes in the topology of the network. In this paper, we propose a solution for the placement of 5G services in dynamic multi-clusters. The learned Deep Reinforcement Learning model is resistant to topology changes (i.e., nodes and links modifications). The proposed model is tested in different network changes scenarios by considering the placement of sets of core network functions (namely, Free5GC core networks) on Kubernetes clusters and is compared to updated models.