基于RapidIO的大容量固态存储系统设计与实现
发布时间:2018-04-25 11:33
本文选题:串行Rapid + IO ; 参考:《电子科技大学》2016年硕士论文
【摘要】:在军事运用领域,由于其对硬件环境要求的严苛性,机械存储设备无法满足需求,绝大部分场合都是使用固态存储技术来实现大容量存储。早期的军用实现方案为基于掉电备份的DRAM存储,但由于该技术需要独立的供电系统,系统功耗高、存储的容价比高等缺点,已逐渐被淘汰,转而由基于闪存的固态存储技术来实现。存储在军事场合的应用,如雷达前端采集数据的存储、传感器前端采集数据的存储等,数据存储的需求大部分是:连续写入;写入、读取带宽需持续保证;系统需具备文件删除等管理功能。面对这些需求,若直接采用固态存储盘(SSD)来进行存储管理,可能会存在多次写入、删除文件操作后,数据存储速度变慢,写入带宽短时间内不满足最低存储要求的情况。为了避免这些问题,目前广泛使用的方法是采用NAND Flash颗粒组成专用存储系统。早期国内各大院校以及科研机构都开发出过基于NAND Flash颗粒的大容量存储设备,但大部分的设计方案还是停留在简单的顺序存储、读取设计上,也就是不支持文件的管理,对多用户接入、多数据源存储、读写同时操作的应用场合是不能够很好的支持。针对日益变化的存储需求,本文设计了一种基于Rapid IO接口的固态存储系统。该存储系统采用同步MLC NAND Flash作为存储介质,使用16通道8级流水写入技术,使后端连续存储速度可达4GB/s以上,且随机1MB块数据读取速度可达2.4GB/s以上。ECC处理方面则采用BCH编解码算法,每个通道使用一个带宽达200MB/s的处理核,从而使整个系统的ECC处理带宽大于2.4GB/s。前端3通道串行Rapid IO接口通过自定义命令通信协议以及数据通信协议,支持多达32个用户同时操作存储系统,以及数据存入的每个物理通道以及逻辑通道独立形成数据文件。缓存方面系统采用高达2GB的DDR3L颗粒组作为数据读写缓存,并且基于LRU技术处理,使数据浏览的响应速度达到最快。最后本文设计了一种容许大碎片的高速存储FTL算法,使本系统在军事应用场合中能够很好的满足文件管理后,存储速度不受影响的要求。
[Abstract]:In the field of military application, due to the harsh requirements of hardware environment, mechanical storage equipment can not meet the demand. In most cases, solid state storage technology is used to achieve large capacity storage. The early military implementation scheme was DRAM storage based on power off backup. However, due to the need of independent power supply system, high power consumption, high storage cost ratio and other shortcomings, this technology has been gradually eliminated. Instead, it is based on flash-based solid-state storage technology. In military applications, such as radar front-end data storage, sensor front-end data acquisition data storage, most of the data storage requirements are: continuous writing, write, read bandwidth need to be continuously guaranteed; The system must have file deletion and other management functions. In the face of these requirements, if the storage management is directly based on solid state storage disk (SSD), there may be many writes, after deleting files, the data storage speed becomes slower, and the write bandwidth does not meet the minimum storage requirements in a short time. In order to avoid these problems, the widely used method is to use NAND Flash particles to form a special storage system. In the early years, universities and scientific research institutes developed large capacity storage devices based on NAND Flash particles, but most of the design schemes remained in simple sequential storage, read design, that is, do not support the management of files. For multi-user access, multi-data source storage, read-write simultaneous operation of applications can not be very good support. A solid state storage system based on Rapid IO interface is designed to meet the changing storage requirements. The storage system uses synchronous MLC NAND Flash as storage medium and 16-channel 8-level pipeline writing technology to make the backend continuous storage speed more than 4GB/s. In the aspect of reading speed of random 1MB block data up to 2.4GB/s, the BCH codec algorithm is adopted. Each channel uses a processing core with bandwidth up to 200MB/s, thus making the ECC processing bandwidth of the whole system more than 2.4 GB / s. The front-end three-channel serial Rapid IO interface supports up to 32 users to operate the storage system simultaneously through the custom command communication protocol and the data communication protocol, and each physical channel and logical channel stored in the data can independently form the data file. In the cache system, the DDR3L granular group up to 2GB is used as the data read / write cache, and based on the LRU technology, the response speed of the data browsing is the fastest. Finally, this paper designs a high-speed storage FTL algorithm to allow large fragments, so that the system can meet the requirements of file management after the storage speed is not affected in military applications.
【学位授予单位】:电子科技大学
【学位级别】:硕士
【学位授予年份】:2016
【分类号】:TP333
【参考文献】
相关期刊论文 前2条
1 华斌;黄杰文;周章伦;孙建涛;张平;;基于FPGA的高速大容量固态存储设备数据ECC的设计与实现[J];科学技术与工程;2010年18期
2 韩茜,罗丰,吴顺君;高速大容量固态存储系统的设计[J];雷达科学与技术;2005年02期
相关博士学位论文 前1条
1 朱岩;基于闪存的星载高速大容量存储技术的研究[D];中国科学院研究生院(空间科学与应用研究中心);2006年
相关硕士学位论文 前8条
1 雷雨;基于PCIE固态硬盘的嵌入式存储技术[D];中国科学院研究生院(光电技术研究所);2014年
2 滕达;基于FPGA的快闪存储器纠错电路设计与实现[D];山东大学;2014年
3 陈仁钢;面向固态存储的BCH-ECC的算法研究与硬件设计[D];重庆大学;2012年
4 张超;高速数据采集与大容量固态存储系统设计[D];西安电子科技大学;2012年
5 李华;大容量固态存储阵列的关键技术[D];西安电子科技大学;2011年
6 吴昊;高速大容量固态存储系统设计[D];西安电子科技大学;2010年
7 雷磊;NAND型FLASH海量存储系统的设计与实现[D];北京理工大学;2008年
8 王s,
本文编号:1801134
本文链接:https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/1801134.html