Research On Flash Translation Layer Based On Data Deduplication Technology

NAND flash has the characteristics like out-of-place updates,erase-before-write and limited number of times of erasure,making NAND flash-based solid state drive(SSD)not well suited for write-intensive applications.Deleting duplicate data is an effective way to reduce pressure on the SSD.Mapping data blocks with the same content to the same physical page reduces the amount of data written to the flash memory.However,the flash translation layer lacks the consideration of the characteristics of the multi-cited physical page.Such a physical page error may cause great destructiveness.In addition,compared with an ordinary physical page,these physical pages are referenced by multiple data blocks and are relatively insusceptible to failure.However,it was repeatedly moved due to the need for garbage collection,which affected the read/write performance of the SSD.In order to solve the above problems,this paper presents the Awareness of the Failure Probability of Multi-Cited Data Flash Translation Layer(AFP-FTL)and the Dedupe-Aware Redundant Strategy(DA-RS).AFP-FTL defines the physical page failure probability of SSD,and manages the data with different failure probability by hierarchical mapping,which can effectively reduce the repeated migration of multi-citation data by garbage collection and improve the SSD performance.According to the damage caused by physical page errors and the probability of errors,DA-RS distinguishes the importance of data and provides dynamic adjustable reliability assurance.In addition,the pre-processing fingerprint calculation reduces the amount of cached data and the overhead of writing data back to flash memory to achieve improved performance.And when the recovery block is selected,the recycling efficiency and the wear of each block are taken into consideration together to further extend the service life of the SSD.It has been proved by experiments that the reading performance of AFP-FTL relative to CAFTL can be improved by 20%-35%,the performance of writing can be increased by 15%-30%,and the erasure times are reduced by nearly 10%,and the erasure distribution is more evenly distributed.Using DA-RS strategy can reduce the physical page uncorrectable error rate of nearly 15%-40%,for partial load can reduce the reading and writing average response time of nearly 10%-15%,and write amplification and CAFTL similar.The new system,which takes into account the characteristics of multiple cited data in depth,ensures the reliability of deduplication systems and reduces the damage caused by data errors.