Advanced Constellation Emulation and Synthetic Datasets Generation for Non-Terrestrial Networks

Mega satellite constellations, now realized entities, encompass thousands of nodes. However, efficient orchestration of multi-hop paths and distributed processing tasks in Non-Terrestrial Networks (NTN) remains a considerable challenge. The integration of NTN systems into 5G cellular networks necessitates innovative adaptations of Software-Defined Networking (SDN) and Multi-access Edge Computing (MEC) to suit the dynamic environments of NTN. In this context, we present MeteorNet, a state-of-the-art emulation tool conceived for satellite constellations. MeteorNet accurately replicates the behavior of NTNs by implementing space orbits, Earth rotation calculations, and Linux network interfaces across diverse network layers. Coupled with a continuous measurement system founded on sFlow, MeteorNet compiles critical switch variables in a centralized database, thus providing a distinctive methodology for creating realistic synthetic datasets. The pertinence of synthetic datasets is paramount in NTN, given the scarcity of operative systems and the inaccessibility of accurate data from the few existing systems due to proprietary constraints. These datasets are instrumental for formulating and training intelligent control algorithms and Machine Learning (ML) models for SDN and MEC advancements in NTN. To illustrate the efficacy of this approach, we explore a realistic networking case study with a ring topology, demonstrating how data models describe intricate routing and edge computing protocols for NTN.