Empowering Satellite IoT for Faster Image Transfers

Abstract

LEO satellite networks are emerging as a global-scale connectivity infrastructure for regions beyond the reach of terrestrial networks. Among them, satellite IoT targets low-power, low-cost IoT applications; however, our real-world measurements reveal that today's commercial satellite IoT still faces substantial challenges in supporting large-volume data transfer for real-world IoT applications. Our results show that a highly-compressed image of only tens of kilobytes typically takes 6–10 hours, extremely exceeding the application time requirements. We find that the bottleneck lies in the direct-to-satellite upload stage, where usable contacts are scarce and underutilized. Moreover, simply adding more nodes does not provide proportional gains, as beacon-triggered upload opportunities remain isolated and exhibit weak correlation. Based on this observation, we propose Co-DtS, a multi-interface upload system that promotes observed beacons into cross-interface coordination signals. Trace-driven evaluation with commercial devices shows that Co-DtS reduces image completion time from 7.62 hours to 0.9 hours.