Wear Leveling Algorithm Research In Solid-State Drive Controller

NAND Flash based SSDs(Solid-State Drives) have been developed rapidly because of apparent advantages over HDDs(Hard Disk Drives), such as consuming less power and are more resistant to harsh environment. Despite of the excellent merits the SSDs hold, they suffer imperfection from the intrinsic characteristics of NAND flash memory. Algorithms referred to as wear leveling for enhancing the endurance and reliability of flash memory are studied in this paper. Data are accessed in different frequencies in flash. Data tended to be updated frequently are called"hot data", whereas relatively infrequently rewritten data are called"cold data". Once hot data keep accessing the same block, the block will be soon worn out and the SSD will be too unstable to use anymore. How to distribute erase operations as evenly as possible around the whole flash memory and lengthen the overall lifespan of SSD is a big concern in the wear leveling algorithm. In this paper, a creative static wear leveling algorithm is proposed to improve the reliability of SSD. Firstly hot blocks and cold blocks are divided into two groups according to their hotness recorded in a hash table, and data swapping operations are applied thereafter. Secondly, by adding hot/cold data identification module based on another hash table in the memory, it can be seen that excellent performance has been achieved with conserved resource. It should be noticed that an approximate dynamic wear leveling module has been applied in our wear levelingmechanism, and the flash system performance is improved.