Monte Carlo Throughput Estimation in Unstable LEO Satellite Networks

TL;DR

This work addresses the challenge of evaluating throughput in unstable LEO satellite networks where inter-satellite links (ISLs) are unreliable. It introduces CAP-uLSN to model time-varying network capacity as a function of ISL availability modeled by a semi-Markov/alternating renewal process, and develops the Monte Carlo Throughput Estimation (MCTE) framework to assess throughput under dynamic traffic and routing. Two MCTE variants are proposed: MCTE-DT for deterministic topology and MCTE-ST for stochastic topology with incremental updates, achieving accurate throughput estimates with lower computational cost than traditional flow-based methods. The approach enables more realistic benchmarking of routing schemes and traffic patterns, and suggests design directions for routing, billing models, and adaptive traffic engineering in future LEO constellations.

Abstract

This study introduces a new framework for analyzing capacity dynamics and throughput performance in Low Earth Orbit satellite networks (LSNs). It focuses on addressing critical gaps in existing models, particularly those concerning unreliable ISLs. Our work systematically resolves two inherent deficiencies in prior research: (1) the conflation of network capacity with maximum throughput, the latter being highly dependent on routing policies and thus failing to reflect the intrinsic characteristics of the system; and (2) the overestimation problem in flow network based throughput calculations, which often generate flow paths that are inconsistent with actual traffic paths. To address these issues, we develop the CAP-uLSN (Capacity under unstable LEO satellites networks) model to characterize time-varying network capacity under stochastic ISL availability. Furthermore, we propose a Monte Carlo Throughput Estimation (MCTE) framework that probabilistically evaluates aggregate throughput performance under dynamic traffic patterns and diverse routing schemes. These insights derived from the CAP-uLSN model and MCTE framework, provide theoretical guidance for optimizing routing schemes (e.g., path selection under throughput fluctuations) and designing adaptive billing models (e.g., distance-based pricing) in future LEO satellite networks.

Figures (9)

• Figure 1: Today's LEO satellite network architecture.
• Figure 2: The ground-satellite relaying traffic and inter-satellite forwarding traffic.
• Figure 3: Alternating renewal process in ISL maintenance. The red line with high (1) and low (0) status denotes ISL availability function $Z_e(t)$.
• Figure 4: Comparison of MCTE-DT and MCTE-ST methods during ISL $e_{ij}$ failure events at time $t_2$. While both MCTE methods recompute $p_1$'s traffic path, MCTE-DT additionally recomputes $p_2$'s route, potentially causing routing oscillations due to changing traffic path. In contrast, MCTE-ST only updates paths affected by the failed ISL ($p_1$ ), maintaining unaffected flows ($p_2$) with their previous routing decisions.
• Figure 5: Ground station distribution, synthetic traffic based on population distributionciesin_gpwv4 and constellation scenario in SNKsnk setups.