Cloud-Edge Resource Management and Migration: Leveraging Online Learning for Digital Twin Re-placement

The increasing demands of devices and services, particularly within the Internet of Everything (IoE), are driving the need for innovative solutions to manage the vast amounts of associated data. Initially used in manufacturing, Digital Twin (DT) technology is now recognized as essential in the 6 G ecosystem, supporting applications like smart cities and autonomous transportation, which benefit from the network's ultr-alow latency and high reliability. However, even if the literature faces the development and management of the DTs, it lacks comprehensive strategies for their deployment and placement in the networks. Current research mainly explores bringing DTs closer to devices through edge frameworks, without addressing dynamic resource management. In this sense, this paper proposes an online-learning-based model for deploying DTs at the edge to meet stringent latency requirements. The conceived approach leverages communication between two entities, a Cloud and an Edge Managers, ensuring optimal DT placement and efficient resource use. A performance evaluation shows the benefits of the conceived solution in terms of convergence time and latency compared to the most used centralized approaches.