Research On Garbage Collection Method That Takes Into Account Wear Leveling For Solid-state Drive

Solid-state drives(SSDs)based on NAND flash memory are praised for their advantages of high bandwidth,high performance,high reliability and low energy consumption.It has been widely used in application scenarios such as personal computers,embedded devices,and data centers.However,flash-based solid-state drives need to perform garbage collection operations to recycle invalid data pages to reuse invalid pages.Normally,garbage collection is very time-consuming,and may sometimes affect the normal read / write operations of the SSD and reduce the performance of the SSD.In addition,since the flash memory block in the SSD has a limit on the number of erasures,SSD also needs to introduce a wear leveling mechanism to avoid excessive wear of some flash memory blocks,and hopes to improve the service life and the reliability of data storage of the SSD through wear leveling methods.Therefore,it is of great practical significance to study the garbage collection and wear leveling problems of SSD.Various garbage collection methods in traditional SSD generally consider the wear leveling and garbage collection separately.However,in actual use,the degree of wear of the flash memory blocks in SSD changes at any time,so it is necessary to consider the balance between wear leveling and garbage collection to obtain the best performance of the SSD.Based on this,this thesis focuses on the in-depth analysis and research on various garbage collection methods and wear leveling problems in SSD,and gives corresponding solutions for the garbage collection method that takes into account wear leveling and the delayed garbage collection algorithm of SSD.The main research work and innovations of this thesis are as follows:(1)A block value calculation method considering wear leveling and garbage collection is given.In the calculation formula of the block value,the efficiency of garbage collection and wear leveling are comprehensively considered,and the actual use of SSD is also considered.In addition,this thesis also introduces the first-input first-output idea,only the blocks written earlier use the block value calculation formula to calculate.That is,the sooner the blocks that have been written are more likely to be selected as victim blocks,which can effectively reduce the number of page migrations,thereby reducing the number of block wear,and better achieve the purpose of garbage collection of SSD that take into account wear leveling.(2)Proposed a garbage collection algorithm that take into account wear leveling(GC-WL).The GC-WL algorithm can automatically adjust the garbage collection algorithm dynamically according to the wear of SSD and the free capacity when selecting the victim blocks in the garbage collection,reducing the number of SSD wear and the overhead of garbage collection,and improving the service life and data reliability of SSD.(3)A delayed garbage collection algorithm(DGC)is proposed.The DGC algorithm cooperates with garbage collection and the event queue in SSD during the garbage collection process.When the read operation in the event queue may conflict with the garbage collection request,the read operation will be prioritized by delaying the garbage collection time,reducing the average read delay time of SSD.(4)A new SSD delayed garbage collection method that take into account wear leveling(DGC-WL)is proposed.The DGC-WL method optimizes both the wear leveling of SSD and the delayed garbage collection at the same time,improves the impact of garbage collection on the performance of SSD,and further improves the service life and data read / write performance of SSD.