Data-driven decision-making framework for optical fronthaul slice resizing in 6G networks

Abstract

The third-generation partnership introduced three main types of slices: enhanced mobile broadband, massive machine-type communication, and ultrareliable low-latency communication. To accommodate these services, the next generation of mobile networks will require architecture with distinct requirements and network slices. To implement these services on an optical fronthaul, the slices will be hosted using lightpaths. Such lightpaths will have to accommodate latency and bandwidth constraints to keep radio units (RUs) and baseband units (BBUs) synchronized. However, the traffic in a slice may vary, and the resources allocated to a long-established lightpath could be out of date, leading to waste or lack of resources. For example, a lack of bandwidth can cause desynchronization between BBUs and RUs. Therefore, the slice must be resized regularly to meet the variable demands. This work proposes a data-driven decision-making (DDDM) framework to resize the fronthaul slices while mitigating the consequences of a lack of bandwidth. The framework uses long short-term memory to implement its analytical stage and integer linear programming (ILP) to reconfigure the entire network when it is required. The results show that the DDDM-based framework outperforms the state-of-the-art ILP-based heuristic by up to 15% in terms of radio blocking mitigation.