Research On New Generation Of Flash Memory Based Solid State Drives Storage System Optimization

In recent years,flash-based solid state drives have been widely used in modern computer systems,including embedded systems,personal computers,and data centers.Compared with traditional hard disk drives,solid state drives have the advantages of low power consumption,high shock resistance,and fast random access performance.However,with the advent of the era of big data and cloud storage,users' demands for storage system performance,data reliability,and lifetime have also become stronger.Therefore,solid state drive storage systems are facing new opportunities and challenges.How to make solid state hard disks provide better storage performance will become one of the main research directions in industry and academia.Currently,flash manufacturers have introduced a variety of new generation of solid state drive design architectures,which have better performance,capacity,and lifetime.However,there are still a series of shortcomings and disadvantages in the new generation of solid state drives.First,based on the basic architecture characteristics of solid state drives,this thesis analyzes their inherent parallel architecture,and uses the parallel architecture to optimize storage performance.Then,a data integrity protection study is conducted to address the problem of data loss due to capacitor aging in the new generation of solid state drives.Finally,this thesis studies the problem of data reprogramming in the 3D solid state drives,and proposes how to use reprogramming to optimize the effective capacity,performance and lifetime of 3D solid state drives.The main research contents of this thesis are as follows:(1)The solid state drive is mainly composed of a multi-level parallel architecture,which can provide a high degree of parallel access,so that access data can be processed on multiple parallel units simultaneously.However,the current access data does not perceive the state of the flash memory chip,which causes an access conflict among the access data that accessing the same parallel unit,resulting in performance degradation.In this regard,this thesis uses the buffer device configured in the modern solid state drives to schedule access data in advance and reallocate access locations,thereby alleviating conflicts among access data on the same parallel unit and improving the performance of solid state drives.(2)The new generation of solid state drives will be configured with a volatile buffer for data temporary storage,thereby improving the performance of solid state drives.To ensure the integrity of data in the buffer,the solid state drive will be equipped with a capacitor in its controller to protect the data in the buffer from power failure.However,the current capacitor devices are facing aging problems,which will reduce their capacitance and make it impossible for all the data in the buffer to be written back to the flash device.Therefore,this thesis proposes to detect the maximum amount of write-back data supported by current capacitor capacitance.Based on this pre-detected value,this thesis considers the current architecture characteristics of the solid state drive to evict the data in the buffer in advance,so as to ensure the optimization of performance while avoiding the data lose problem caused by capacitor aging.(3)3D solid state drives can provide better storage capacity for storage systems,but because the stacking structure of 3D solid state drives cannot be increased indefinitely,this thesis will study how to further increase the effective capacity of 3D solid state drives while achieving better lifetime and performance.In view of the above problems,this thesis proposes to program a 3D TLC multiple times to improve the storage capacity,thereby reducing the number of garbage collections,and finally to optimize the lifetime and performance of 3D solid state drives.In order to realize the reprogramming of 3D TLC flash memory cells,the voltage distribution state is redistributed,and the voltage distribution is defined according to the current reprogramming times.The 3D TLC contains eight voltage distribution states.Through the redistribution of the voltage states,the eight voltage distribution states can be reprogrammed twice.By implementing this reprogramming scheme in a flash array,we have greatly reduced the need for flash storage space for data access and reduced overall garbage collection times.Aiming at the above-mentioned optimization research techniques,this thesis uses a solid state drive simulator and an industrial-grade flash chip test platform for experiments.The experimental results prove that the above research technology can effectively improve the performance and lifetime of solid state drives.This thesis fully considers the inherent characteristics of solid state drives,and realizes the overall system optimization research of new-generation solid state drives from the perspective of basic parallel architecture of solid state drives,data integrity protection,performance and lifetime optimization,and further promotes the new-generation storage devices in big data and cloud storage era.