eMMC可靠性分析与增强对策研究

eMMC是可以自我管理的闪存（NAND flash）芯片，内部集成了控制器，实现了多芯片封装，简化了应用接口设计，节省了板级空间，在移动终端领域得到了广泛应用。然而，基于flash的存储器件在掉电时容易出现比特位损坏、元数据损坏、非串行写、甚至器件不可用等可靠性问题。针对如何提高eMMC的可靠性进行了研究，提出了一些可靠性增强措施。首先，全面梳理了eMMC规范提供的可靠性特性，包括分区管理、写保护管理、可靠写、数据标签、断电通知等。继而对现有的各种可靠性措施进行了综合分析，并提出了一些提升可靠性的措施，包括，文件系统感知的磨损均衡、根据数据属性映射到合理分区、减少写入次数降低写放大效应，面向手机应用的元数据去冗、应用差异化细粒度控制可靠性、确保一致性的前提下减少日志等。

Analyzing and enhancing the reliability of eMMC devices

eMMC device is self-managed flash memory storage chip with a controller to achieve multi-chip package, simplify the interface and save board space, thus eMMC is widely used in mobile devices. However, for those flash-based storage devices, many reliability problems including bit corruption, metadata corruption, unserializable writes or dead device are likely to occur in the presence of power failure or system crash. This paper studied how to improve the reliability of eMMC devices and presented several solutions to enhance reliability. Firstly, it made a comprehensive analysis on reliability features of eMMC specification, including made a number of research proposals, including partition management, write protect, reliable write, data tag, power off notification. Based on the analysis of reliable features, we proposed several solutions, including file system aware wear leveling, data attribute aware partition mapping, reducing writes, reducing metadata redundancy, application aware fine-grained reliability control, and reducing journaling.