On the queuing delay of time-varying channels in Low Earth Orbit satellite constellations

Abstract

Low Earth Orbit (LEO) satellite constellations are envisioned as a complementary or integrated part of 5G and future 6G networks for broadband or massive access, given their capabilities of full Earth coverage in inaccessible or very isolated environments. Although the queuing and end-to-end delays of such networks have been analyzed for channels with fixed statistics, currently there is a lack in understanding the effects of more realistic time-varying channels for traffic aggregation across such networks. Therefore, in this work we propose a queuing model for LEO constellation-based networks that captures the inherent variability of realistic satellite channels, where ground-to-satellite/satellite-to-ground links may present extremely poor connection periods due to the Land Mobile Satellite (LMS) channel. We verify the validity of our model with an extensive event-driven simulator framework analysis capturing the characteristics of the considered scenario. We later study the queuing and end-to-end delay distributions under such channels with various link, traffic, packet and background conditions, while observing good match between theory and simulation. Our results show that ground-to-satellite/satellite-to-ground links and background traffic have a much stronger impact over the end-to-end delay in mean and particularly variance, even with moderate queues, than unobstructed inter-satellite connections in outer space on an established path between two ground stations and through the constellation. This might hinder the usability of these networks for services with stringent time requirements.