Provisioning Anycast Connections in Traffic-Prediction-Assisted Multilayer Networks

Anycast, as a one-to-one-of-many communication strategy directing user requests to the nearest or best replica of content stored across multiple data centers (DCs), has gained widespread adoption for improving network performance by reducing latency, and enhancing resilience. However, the growing reliance on anycast in large-scale networks has introduced new challenges, particularly in terms of congestion on the links connecting (DCs), which can degrade network performance and reliability. This paper explores the impact of provisioning anycast connections on multilayer backbone optical networks, particularly under time-varying traffic conditions. We investigate the effects of varying anycast traffic proportions and the number of (DCs) on network performance and resource utilization. Our findings demonstrate the potential of a traffic-prediction-based algorithm to optimize anycast provisioning, contributing to more efficient network management and providing a foundation for future research in dynamic resource allocation and multilayer network design.