A Mobility-Aware Reinforcement Learning Proactive Solution for State Data Migration in Edge Computing
Elias Dritsas, Kostas Ramantas, Christos Verikoukis · 2024
To handle the demands of modern applications for storage, computing, low latency, and bandwidth, various services are offloaded from the cloud to edge servers, bringing them closer to end-users. This shift in computing resources closer to the users is known as Edge Computing whose services are highly affected by users' mobility. As users move and change geographical locations, the services provided by an edge server may be compromised or become distant from the user, leading to a decrease in the quality of service (QoS) for applications. Service migration is a common approach to ensure that essential services remain close to users, thereby improving the application QoS. The synergy of Edge Computing with Artificial Intelligence (AI) can open up new capabilities for the Lifecycle Management (LCM) of applications, besides the adoption of the innovative microservice-based architecture for building them and, especially, stateful to ensure seamless service continuity and minimal latency during user's mobility. The leverage of AI-based LCM of such applications can significantly benefit edge computing by automating and optimising, in terms of resource allocation, the process of decision-making during deployment and run-time operation of microservices running on edge devices. In this direction, we elaborate a mobility-induced proactive migration mechanism empowered by efficient Deep Learning (DL) and Reinforcement Learning (RL) algorithms to identify the optimal edge node for application and state data migration without violating Service Level Agreement - SLA (e.g., latency) and ensure efficient resource usage at the edge.