Double Replication MDS Codes for Wireless D2D Distributed Storage Networks
Juan Li, Shushi Gu, Ye Wang, Qinyu Zhang · 2018
Caching files between base station and mobile devices by distributed storage system (DSS) is one of effective measures to save energy and bandwidth cost lately. In this paper, we consider a novel wireless network where devices use device-to-device (D2D) communication to transmit and storage content. Several schemes, such as maximum distance sparable (MDS) code and regenerating code, are employed and analysed for downloading and repairing content when nodes left and data lost in network. We found that MBR is not suitable for D2D-DSS networks due to higher download cost, surprisingly, but MDS is a convenient and appropriate scheme for majority practical D2D networks. However, MDS has to transfer the entire file size of the data to repair just a single node, which brought larger repair bandwidth. For this problem, we propose a new caching scheme, termed double replication MDS (DR-MDS) codes, utilize more storage nodes to save more repairing cost and reduce download cost simultaneously. We derive and simulate the communication cost performance of DR-MDS compared to traditional DSS coding schemes, shown that DR-MDS scheme significantly outperforms that of traditional MDS schemes and replication, that prove the practicability and feasibility of DRMDS scheme used in D2D distributed storage networks.