Measurement Analysis And Optimization Of Progressive Wear Leveling Algorithm For NAND Flash

With the rapid development of embedded and storage technology,NAND flash memory has not only been widely used in various electronic devices,but has become an important medium for data storage in embedded systems.Compared with mechanical magnetic disks,solid-state drives uses flash memory as the primary storage medium,due to its non-volatility,small size,low power consumption.However,since erase count of flash blocks is limited,once the erase count reaches the upper limit,the data stored in the block is not reliable.Therefore,the corresponding file management system is proposed to improve the performance of the solid-state drive.The wear leveling algorithm can effectively equalize erase count of flash blocks,which cause all flash blocks to reach the upper limit of erase count simultaneously,thereby extending the lifetime of flash memory.However,the existing wear leveling algorithm with better performance generally suffers from the problem of extra loss.Due to the inability to adapt well to the working environment with more cold data,the algorithm makes the write amplification larger,which affect the performance of the solid-state disk and reduce the flash service lifetime.To address this issue,the wear leveling algorithm needs to reduce the number of internal data migrations.This dissertation based on the traditional wear leveling algorithm,mainly studies the flash-memory characterization,and optimizes the threshold for triggering wear leveling mechanism and the selection conditions of victim block.Furthermore,the simulation compares the performance of the two algorithms suffering from different ratios environments of hot and cold data.The main research contents and innovations are as follows:(1)Firstly,this paper studies the design principle of traditional wear leveling algorithm and further analyzes its advantages and disadvantages.Then,the hot and cold data identification and data migration conditions are studied.Finally,the performance and lifetime calculation formula of the algorithm are analyzed in detail,which provides a theoretical foundation for the performance comparison of the algorithms.(2)This paper studies the flash-memory characterization of solid-state drives and tests the program disturb and operation time of flash memory.Then,this paper further performs bad block processing on solid-state drives,and studies the importance of bad block management mechanism by analyzing data stored in different types of bad block.Moreover,the impact of bad block management on the performance and reliability of flash memory is studied.(3)In order to overcome the problem of the extra cost of the progressive wear leveling algorithm(PWL),a dynamic threshold wear leveling algorithm based on the upper limit of erase count is proposed.In different ratios environments of hot and cold data,the performance of the proposed algorithm and PWL algorithm are compared by simulation.The results show that in a large number of cold data environments,the proposed algorithm not only reduces the wear loss,but also effectively extend the lifetime of the flash memory.