关键词: 云存储, 网络编码, 单位向量, 冗余, 每字符运算次数

Abstract:

According to structural characteristics and availability requirements of cloud storage, a semi-random linear network coding (SRLNC) algorithm is proposed for data redundancy. The data block is split into pieces, encoded by coding matrix composed of unit row vectors and random row vectors on finite fields GF(2^s), stored in data nodes dispersedly. In decoding process, the data block can be restored with the strategy of unit row vectors priority. The probability of unique solution is analyzed for SRLNC decoding equations, a new operations per character index is defined to measure algorithm efficiency. Experiments prove that the operating time of the decoding process in the proposed algorithm exponentially reduces with the increase of the node availability. When the availability of each node≥0.8 and redundancy≤3, the encoding operation of the proposed algorithm is faster up to 33%, and decoding operation of the proposed is 5 times faster than that of random linear network coding algorithm, respectively. It is shown that the proposed algorithm is applicable to "write once read many" cloud storage system.

Key words: cloud storage, network coding, unit vector, redundancy, operations per character