Digital Twin Enriched Green Topology Discovery for Next Generation Core Networks

Abstract

Topology discovery is the key function of core network management since it utilizes the perception of data and mapping network devices. Nevertheless, it holds operational and resource efficiency complexities. For example, traditional discovery cannot perform predictive analysis to learn the network behavior. Moreover, traditional discovery periodically visits IP ports without considering the utilization levels, which leads to high resource usage and energy consumption. Hence, it is necessary to integrate intelligent methods into traditional discovery to deeply understand the behavioral pattern of a core network and recommend action to avoid these intrinsic complexities. Therefore, we propose a Digital Twin (DT) enriched Green Discovery Policy (DT-GDP) to serve a green discovery by using the increased intelligence and seamless assistance of DT. DT-GDP jointly uses the outputs of two modules to calculate the total energy consumption in Watts. In the energy module, we consider the service power, idle state power, and the cooling power of an IP port and derive a novel energy formula. In the visit decision module, we use Multilayer Perceptron (MLP) to classify the IP ports and recommend visit action. According to experimental results, we achieve a significant reduction in the visited ports by 53% and energy consumption by 66%.