An Efficient Parallel Simulation Architecture for Ultra-Mega-Scale LEO Satellite Networks

As Low Earth Orbit (LEO) satellite constellations continue to expand in scale, traditional satellite network simulation approaches will face limitations in performance and scalability when simulating large-scale satellite networks. To address this issue, we propose an efficient parallel simulation architecture named HyperLEO, specifically designed for ultra-large-scale LEO satellite networks. The architecture leverages MPI-based parallelization techniques to significantly enhance the simulation efficiency. We further introduce a parallel discrete event simulation (PDES) synchronization algorithm tailored to the dynamic characteristics of satellite networks and develop a load-balanced partitioning algorithm. Experimental results show that HyperLEO achieves up to a $19.7 \times$ speedup using 32 processes over serial simulations at the LEO-2500 scale and is capable of supporting simulations of constellations with up to 20,000 satellites. Our work provides an efficient and scalable approach for protocol development and performance evaluation in ultra-large-scale LEO satellite networks.