Deep Predictive User Path-Inclusive Service Mobility in B5G/6G Edge Clouds

Multi-access Edge Computing (MEC)-enabled Beyond 5G (B5G) or 6G networks require service migration to maintain service continuity and acceptable quality of experience for high-mobility User Equipment (UE). This involves migration of containerized applications to between Edge Hosts via the backhaul network as the associated UE moves away from its current gNodeB. Although crucial for both UEs and mobile network operators (MNOs), frequent migration not only prolongs service disruption time (downtime) for UEs but also increases operational expenses for the MNO. To address this problem, our study proposes a Deep Predictive User Path-Inclusive Service Mobility (DUPSM) approach for efficient management of the end-to-end migration process including deciding when and where to migrate the service. It features a deep Actor-Critic agent that learns suitable decision policy to balance cost-benefit trade-off interms of migration cost and quality of experience, respectively. As user mobility largely influences migration decisions, the proposed method incorporates user mobility path forecasting through a Generative Adversarial Network (GAN) based prediction technique to enhance ability for selecting suitable Edge Host while considering resource variations. The effectiveness of the proposed method is assessed through simulation experiments, showcasing an average reduction in service downtime of 22% when compared to a baseline scheme.