基于软硬件协同设计的闪存存储系统关键技术研究与实现
发布时间：2018-09-06 17:23
【摘要】：随着移动设备、社交网络、物联网等多种应用的蓬勃发展,人类社会产生的数据呈爆炸式增长。以海量数据为基础的数据密集型应用的出现,不仅对存储系统的容量提出了更高的要求,还在带宽、延迟、功耗、可靠性等方面提出了更高的要求。由于磁盘机械装置的限制,基于磁盘的存储系统已无法满足大规模数据处理在延迟、带宽、功耗等方面的需求。闪存以其高性能、非易失、低功耗、耐振动等特点已被广泛部署在高端存储系统中。另外,文件系统包含更多的语义信息,在设计闪存存储体控制器时,如果能充分利用文件系统的信息,将更好地提高存储系统的性能。本文针对闪存存储系统,通过文件系统与闪存存储体控制器协同设计,对控制器的映射表、通道并行和热点数据管理进行优化。本文主要工作及创新点如下:1.研究了闪存存储体控制器的功能架构,设计实现了闪存芯片同步操作控制器模块,闪存转换层(FTL)模块和热点数据管理模块。2.针对如何有效利用DRAM空间问题,提出了基于协同设计的映射表优化方法,在闪存控制器层实现块映射,文件系统层实现页粒度管理。减少映射表的同时保持存储系统高性能。3.针对传统闪存存储体通道分配策略简单的问题,提出了基于协同设计的通道并行优化方法,闪存控制器层实现指定通道写功能,文件系统层实现通道分配功能,并以此为基础实现了三种通道分配策略,提高通道并行性。4.针对传统闪存存储体热点数据识别精度不高且占用存储体控制器资源的问题,提出了基于协同设计的热点数据管理优化方法,充分利用文件系统的信息进行热点数据识别,控制器层利用文件系统下发的热点语义进行缓存管理,提高闪存存储体控制器的缓存命中率。
[Abstract]:With the rapid development of mobile devices, social networks, Internet of things and other applications, the data generated by human society is exploding. The emergence of data-intensive applications based on mass data not only requires higher capacity of storage system, but also requires higher bandwidth, delay, power consumption, reliability and so on. Due to the limitation of disk mechanical devices, the disk based storage system can no longer meet the requirements of large-scale data processing in terms of delay, bandwidth, power consumption and so on. Flash memory has been widely used in high-end memory systems because of its high performance, non-volatile, low power consumption, vibration resistance and so on. In addition, the file system contains more semantic information. In the design of flash memory controller, if the file system information can be fully utilized, the performance of the storage system will be improved better. In this paper, the mapping table, channel parallelism and hot data management of the controller are optimized through the collaborative design of the file system and the flash memory controller for the flash memory storage system. The main work and innovation of this paper are as follows: 1. The function architecture of flash memory controller is studied, and the synchronous operation controller module of flash memory chip, the (FTL) module of flash memory conversion layer and the hot data management module. 2 are designed and implemented. Aiming at the problem of how to utilize DRAM space effectively, this paper presents an optimization method of mapping table based on collaborative design, which implements block mapping in flash controller layer and page granularity management in file system layer. Reduce mapping table while keeping storage system high performance. 3. Aiming at the problem of simple channel allocation strategy of traditional flash memory, a parallel channel optimization method based on collaborative design is proposed. Flash memory controller layer realizes channel write function and file system layer realizes channel allocation function. On this basis, three channel allocation strategies are implemented to improve channel parallelism. Aiming at the problem that the recognition accuracy of traditional flash memory hot spot data is not high and the storage controller resources are occupied, an optimization method of hot spot data management based on collaborative design is proposed, which makes full use of the information of file system to identify hot spot data. In order to improve the cache hit rate of flash memory controller, the controller layer uses the hot spot semantics of file system to manage the cache.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333
，
本文编号：2227033
[Abstract]:With the rapid development of mobile devices, social networks, Internet of things and other applications, the data generated by human society is exploding. The emergence of data-intensive applications based on mass data not only requires higher capacity of storage system, but also requires higher bandwidth, delay, power consumption, reliability and so on. Due to the limitation of disk mechanical devices, the disk based storage system can no longer meet the requirements of large-scale data processing in terms of delay, bandwidth, power consumption and so on. Flash memory has been widely used in high-end memory systems because of its high performance, non-volatile, low power consumption, vibration resistance and so on. In addition, the file system contains more semantic information. In the design of flash memory controller, if the file system information can be fully utilized, the performance of the storage system will be improved better. In this paper, the mapping table, channel parallelism and hot data management of the controller are optimized through the collaborative design of the file system and the flash memory controller for the flash memory storage system. The main work and innovation of this paper are as follows: 1. The function architecture of flash memory controller is studied, and the synchronous operation controller module of flash memory chip, the (FTL) module of flash memory conversion layer and the hot data management module. 2 are designed and implemented. Aiming at the problem of how to utilize DRAM space effectively, this paper presents an optimization method of mapping table based on collaborative design, which implements block mapping in flash controller layer and page granularity management in file system layer. Reduce mapping table while keeping storage system high performance. 3. Aiming at the problem of simple channel allocation strategy of traditional flash memory, a parallel channel optimization method based on collaborative design is proposed. Flash memory controller layer realizes channel write function and file system layer realizes channel allocation function. On this basis, three channel allocation strategies are implemented to improve channel parallelism. Aiming at the problem that the recognition accuracy of traditional flash memory hot spot data is not high and the storage controller resources are occupied, an optimization method of hot spot data management based on collaborative design is proposed, which makes full use of the information of file system to identify hot spot data. In order to improve the cache hit rate of flash memory controller, the controller layer uses the hot spot semantics of file system to manage the cache.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333
，
本文编号：2227033
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2227033.html
