Fast-Robust Aggregation and Multicast Coded Data Recovery for LEO Satellite Network

The Low-Earth Orbit (LEO) satellite network (LSN) serves as a critical platform for managing data, such as remote sensing images, leveraging its capabilities in on-board processing and inter-satellite links. However, due to factors like space radiation and battery depletion, multiple satellite nodes may fail, resulting in the loss of data and ultimately leading to the unavailability of LSN. Although Erasure Coding (EC) ensures data reliability with low storage overhead, it incurs massive recovery traffic, affecting the efficiency. To address this challenge, we introduce the aggregation and multicast coded recovery (AMCR) scheme, which utilizes Reed-Solomon (RS) codes to achieve fast and robust simultaneous recovery of lost data across multiple failed nodes. First, we propose the multi-weight staged recovery graph (MWSRG) model, i.e., a staged graph with edge weights, to measure various factors affecting recovery performance, including propagation delay, transmission delay, and energy cost. Second, we design two algorithms for single stripe recovery, which reduce the energy cost while maintaining a propagation-delay bound. The aggregation-based multiple single node recovery trees construction (A-MSRT) constructs single-node recovery trees from propagation-delay shortest paths, and the multicast-based aggregation of multiple recovery trees (M-AMRT) uses the recovery tree with the longest delay as the initial one in the output of A-MSRT, adds remaining replacement nodes, and greedily reduces energy cost without exceeding the initial tree's propagation-delay bound. Additionally, for multi-stripe recovery, we further propose the pipeline aggregation and multicast coded recovery (PAMCR) mechanism, which further reduces recovery delay by decentralizing the stripe synchronization process to aggregation nodes. Simulation results show that under different network conditions in LSN, AMCR outperforms other schemes in terms of recovery delay and its energy cost is 5.6% lower than that of DE-MWPT, while PAMCR can further reduce the recovery delay by 36.4% compared to AMCR.