Research And Design Of Wear Leveling Algorithm Based On Solid State Drives

With the advent of the information age,the performance of memorizer is more and more important in the storage and processing of mass data;traditional mechanical hard drives based on Winchester technology are constrained by their physical structure and have been unable to meet the growing needs of the people.A solid state drive which uses semiconductor as its storage medium has gradually entered people's field of vision as a new storage device;solid-state hard drive has the characteristics of high read and write speed,low power consumption,strong shock resistance and a wide operating temperature range,are highly concerned by domestic and foreign researchers.With the progress of the manufacturing process and the improvement of the cost performance,solid-state hard drive gradually showing the trend to replace the traditional mechanical hard disk.However,due to the inherent defects of flash memory,the blocks composed the flash typically have a limited number of erase cycles;once reached its erasing limit,the data on the storage blocks will become unreliable.Therefore,in order to achieve the purpose of extending the life of solid-state hard drive,we need to use the wear leveling algorithm to evenly distribute the erase operation to all blocks to ensure that all blocks reach its erasure limit at the same time.In this paper,the basic principle of static wear leveling algorithm,the research status and the main factors influencing its performance are analyzed in detail,and two different wear leveling algorithms are proposed.The main works and innovations of this paper are as follows:1.A progressive group-based wear-leveling algorithm is proposed.The algorithm can notably reduce the consumption of RAM space required by wear-leveling algorithm while maintaining the wear-leveling performance.In addition,the algorithm also provides a new way to identify the cold data more accurately which dramatically minimize the extra erase operation due to data migration.The experimental results of the simulation through the SSDsim reveal that our proposed algorithm consumes only 8%of memory space compared to the classical deterministic algorithm while showing roughly the same wear-leveling performance.2.A new scheme for hot and cold data recognition based on Multiple Advanced Bloom Filter is proposed,and based on it a wear leveling algorithm based on hot and cold data separation is proposed.By improving the writing mode of the Bloom filter,the false positive of the Bloom filter is greatly reduced without changing the size of the Bloom filter,and the accuracy of the identification of the hot and cold data is improved.By separating the hot and cold data make that the data having the same attribute are allocated in the same block so that the change of the data is synchronized,which greatly reducing the number of valid page transfers in the garbage collection operation.At the mean time,the cold data are assigned to the "old blocks" that have severely worn and the hot data are assigned to the "young blocks" of slightly wear to reduce the probability of triggering static wear leveling algorithm and decrease the additional writes caused by the wear leveling operation.The validity of the algorithm is proved by experimental simulation and theoretical analysis.