基于网络编码的多媒体云存储系统单节点修复算法

Research on single node repair algorithm for multimedia cloud storage system based on network coding

针对现有多媒体云存储系统单节点修复算法中,有限域大,编解码复杂度高,磁盘I/O开销大,存储开销与修复带宽开销不均衡等问题,基于网络编码提出一种多媒体云存储系统单节点修复算法.该算法将系统中节点已分组的多媒体文件数据进行分组存储,并将组内数据在GF(2)有限域上进行异或编码,当部分节点失效时,新生节点只要连接同组中2~3个未失效节点即可精确修复失效节点中的数据.理论分析与仿真实验结果表明:该算法可减小编解码与修复的复杂度,降低磁盘I/O开销;在相同条件下,该算法存储开销与基于最小存储再生码(minimum storage regeneration codes,MSR)算法的存储开销一致,而修复带宽开销与最小带宽再生码(minimum bandwidth regeneration codes,MBR)算法接近.

Concerning that the existing single node repair algorithm for multimedia cloud storage system has problems such as big finite field,high coding and decoding complexity,high cost of disk I/O,unbalance of storage cost and repair bandwidth cost,this paper proposes a single node repair algorithm for multimedia cloud storage system based on network coding.The algorithm groups the nodes in the system,and the multimedia file data are stored in packet storage.The data in groups are coded in GF(2)finite field with XOR.Therefore,when a node fails,connect only 2 or 3 nodes in the same group,the data in failure node can be precisely repaired.Theoretical analysis and simulation results show that the algorithm in this paper can reduce coding and decoding complexity and repair complexity as well as the cost of disk I/O.Under the same conditions,this algorithm has the same storage cost with the algorithm based on MSR,and has a lower repair bandwidth cost,which is close to the repair bandwidth of the algorithm based on MBR.