Storage-Transmission Integrated Erasure Coding Aggregation Repair Techniques in LEO Satellite Data Centers

Zhikai Zhang, Qinyu Zhang, Shushi Gu, Jinbei Zhang, Wei Xiang

Abstract: The LEO satellite data center (LSDC) is becoming a key application paradigm for the future satellite Internet with global information fusion, which consists of numerous satellite storage nodes forming distributed storage systems. Satellite on-board disks inevitably fail due to space radiation and battery depletion, resulting in data loss. Erasure coding (EC) can guarantee data reliability at the expense of heavy repair traffic. To mitigate this challenge, multiple repair data flows can be aggregated progressively into a single data flow during transmission. In this paper, we employ storage-transmission integration to LSDC and propose the erasure coding aggregation repair (ECAR). Firstly, ECAR models repair data flows in LSDC as a staged tree graph, called multi-weight path tree (MWPT). The edge weights in MWPT are metricized by multiple factors affecting the repair performance, including transmission time, propagation delay and energy cost. Then, we give a delay-sensitive energy-efficient MWPT (DE-MWPT) construction method for single stripe repair, which reduces the energy cost while guaranteeing the minimum repair delay. We also propose the pipeline erasure coding aggregation repair (PECAR) for multiple stripe repair, which further reduces the repair delay by decentralizing the stripe synchronization process to aggregation nodes. Finally, we design a storage-transmission integrated simulation platform for LSDC and deploy the prototype of MWPT in it. The experimental results show that DE-MWPT can reduce the repair delay by 7.3% and energy cost by 6% compared with the existing aggregation repair policies, while PECAR can further reduce the repair delay by 43.9% compared to ECAR.

External IDs:dblp:journals/tvt/ZhangZGZX24