Research On Redundancy And Security Of Distributed Storage System Based On Tornado Code

Along with the increased information, the need of data availability is also increasing. The simple storage system and traditional methods of redundancy can not match the need. The erasure code that play an important role in the communication have applied broad because of the significant performance.LDPC code has two notations of sparse matrix and bipartite graph. According to the different distribution of degrees, LDPC code is separated to regular degree code and irregular degree code. Tornado code is the special irregular degree LDPC code, has the characters of cascade irregular degree and the operation of encode and decode only with the exclusive-or (XOR), discuss the creation algorithm of bipartite graph and the process of encode and decode, and prove that the Tornado code is better than the regular degree code in theory.Data redundancy can improve the date reliability. The cost of storage in the redundancy storage system based on replica is larger, the ability of extension in the redundancy storage system based on RAID is weaker and the expense of computation in the redundancy storage system based on RS code is larger, so the distributed redundancy storage system based on Tornado code is suggested according to the analysis. The whole system is analyzed from the four aspects of failure detection, read and write, recovery and garbage collection. At the same time, compare the redundancy storage system based on replica with the redundancy storage system based on erasure code in theory, finally conclude that in the same condition the redundancy storage system based on erasure code have better data availability, less storage overhead and longer repair epoch.In order to solve the intrusion to the redundancy storage system, the secure distributed redundancy storage system based on Tornado code should be suggested. The character of hash function is hardly forged, first the block of data in the distributed system is encoded by Tornado code, then compute the value of hash function as the flag, and improve the security of system through judging the flag. Random times of hash function and the cascade of the hash value of data blocks are provided to compensate the defection of hash function, improve the security further.