闪存文件系统UBIFS的分析与优化

发布时间：2018-03-27 08:23

本文选题：UBIFS　切入点：NAND型闪存　出处：《南京大学》2013年硕士论文

【摘要】：闪存(Flash Memory)是一种常用于嵌入式系统的外存储器,具有非易失、抗震荡、低功耗、体积小等优点。闪存主要包括两种类型：NAND型和NOR型。近年来,随着NAND型闪存容量的快速增长,性价比的不断提高,它比NOR型闪存更加广泛地应用于日常消费电子产品中。闪存的普及不但取决于物理特性的优劣,还取决于闪存文件系统的数据管理性能。传统的闪存文件系统已经不能满足用户在大容量NAND型闪存环境下对挂载速度、内存消耗、读写速度、磨损均衡等指标的要求。作为新一代的闪存文件系统,UBIFS是在Linux环境下实现的日志结构文件系统,解决了NAND型闪存作为MTD设备使用时遇到的若干技术瓶颈,具有广泛的应用前景。本文展开了UBIFS的分析研究工作,并在快速挂载和热数据识别方面设计了两种优化方案,改进后的UBIFS达到了预定的优化目标,具有一定的学术价值。目前有关UBIFS的中文分析文献数量很少,对UBIFS文件系统的详细分析就显得有必要。本文概括介绍了UBIFS的系统结构,并且从UBI模块和UBIFS模块两个角度出发对UBIFS的设计原理进行了阐述,分析了映射管理、磨损均衡和文件索引等关键技术。针对分析过程中发现的问题,本文分别在两个层次上提出了优化方案。在UBI模块层,提出了一种以关机速度变慢为代价换取开机速度提高的优化方案。随着闪存容量的不断增大,挂载UBI模块时扫描所有物理块头部信息的时间也不断增加,导致嵌入式系统的启动变慢,这限制了UBIFS在大容量闪存领域的应用前景。UBI模块卸载时将物理块的头部信息集中保存到预留块中,下次挂载时直接从预留块中读取头部信息,这样可以减少物理块的扫描时间,从而加快开机速度。测试结果表明,该优化方案可以显著提高UBI模块的挂载速度。在UBIFS模块层,写入数据被分为三种类型,不同类型的数据采用不同的物理块分配策略。但是,这种数据分类方法是静态的,无法动态识别用户数据中的热数据,从而影响了磨损均衡性能。通过借鉴内存页面置换算法,本文提出了两级LRU(Least Recently Used)算法,并作为热数据识别算法引入到UBIFS中,使数据的分类更加准确。经过实验验证,两级LRU算法可以提高UBIFS文件系统的磨损均衡性能。
[Abstract]:Flash memory (Flash memory) is a kind of external memory commonly used in embedded system. It has the advantages of non-volatile, anti-shock, low power consumption and small volume. Flash memory mainly includes two types: NAND type and NOR type. In recent years, with the rapid growth of NAND flash memory capacity, With the increasing performance and price ratio, flash memory is more widely used in daily consumer electronics than NOR flash memory. The popularity of flash memory depends not only on physical properties, but also on physical properties. It also depends on the data management performance of flash file system. The traditional flash file system can no longer meet the user's load speed, memory consumption, read and write speed in large capacity NAND flash memory environment. As a new generation flash memory file system, UBIFS is a log structure file system implemented in Linux environment, which solves some technical bottlenecks when NAND flash memory is used as a MTD device. In this paper, the analysis and research work of UBIFS is carried out, and two optimization schemes are designed in terms of fast mount and thermal data recognition. The improved UBIFS achieves the target of optimization. It is necessary to analyze the UBIFS file system in detail. This paper introduces the system structure of UBIFS. The design principle of UBIFS is expounded from two aspects of UBI module and UBIFS module, and the key technologies such as mapping management, wear equalization and file index are analyzed. In the UBI module layer, an optimization scheme is proposed in exchange for the increase of startup speed at the cost of slower shutdown speed. With the increasing of flash memory capacity, the optimization scheme is proposed in this paper. The amount of time it takes to scan all physical block header information while mounting a UBI module is also increasing, resulting in slower startup of the embedded system. This limits the application prospects of UBIFS in the field of mass flash memory. When the UBI module unloads, the header information of the physical block is stored in the reserved block, and the header information is read directly from the reserved block on the next mount, which can reduce the scanning time of the physical block. The test results show that the optimized scheme can significantly improve the mount speed of UBI module. In the UBIFS module layer, the write data is divided into three types, and different types of data are allocated by different physical blocks. This data classification method is static and can not dynamically identify the thermal data in user data, thus affecting the performance of wear balance. By using the memory page replacement algorithm for reference, this paper proposes a two-level LRU(Least Recently used algorithm. As a thermal data recognition algorithm, it is introduced into UBIFS to make the classification of data more accurate. Experimental results show that the two-level LRU algorithm can improve the wear equalization performance of UBIFS file system.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【参考文献】