2017 年 14 巻 24 号 p. 20171131
To recover the errors occurred in flash-based storage, well-known error correction codes (ECC) are integrated into flash memory controller. However, the existing error correction codes have their inherent limits to recover more error than they recover, so more additional schemes should be considered for the higher reliable flash storage systems. In this paper, we investigate duplicated data management to enhance data reliability as replication. To minimize space overhead of the duplication management, we consider two issues, minimizing fingerprint length and enlarging hash chunk size of duplication checking. At most cases, the proposed duplication management scheme can have better reliability than adding more ECC parity bits in the aspect of space overhead.