Cost Optimal Regenerating Codes Design for Satellite Clustered Distributed Storage System

Abstract

Satellite clustered distributed storage system (SCDSS) is a space-ground integrated network architecture which can provide global wide-coverage and inter-connections among data centers around world. Some soft errors caused by single event upsets (SEUs) in outer space make many failures of on-orbit storage devices, which causes stored data lost. In consequence, effective storage codes, e.g., regenerating codes(RCs) and generalized regenerating codes(GRCs), are crucial to ensure the availability of data with the objective of minimizing transmission cost when repairing the failure nodes. However, heterogeneity is an inevitable problem that result in high cost, due to the diversity of communication links with different transmit powers, link bandwidths and communication distances. Existing coding strategies can not provide an effective solution to deal with the heterogeneity of bandwidth costs among different satellite clusters. In this paper, we propose an asymmetric repair strategy of RCs and GRCs, and derive the upper bound of achievable capacity for data storage in heterogeneous clusters. Then, based on the satellite link analysis, we define cost coefficient, and give the expression of data repair cost, which can be optimized by solving a linear programming problem. Finally, numerical results demonstrate that, compared with the repair cost of typical coding strategies of RCs and GRCs, our asymmetric repair strategy performs better performance in reducing the repair cost, as an effective solution for data availability maintenance in SCDSS.