【作者】 朱春霖；

【导师】 冯丹；

【作者基本信息】 华中科技大学 ， 计算机系统结构， 2008， 硕士

【摘要】 随着个人电脑市场的增长和存储设备数据存储能力的不断提高,个人数据量变得越来越大。文件系统作为管理用户数据的基础,其名字空间变得越来越复杂,这导致文件系统元数据访问性能变得越来越低。文件系统名字空间管理对于文件系统性能表现至关重要。从面向应用的角度,提出了一种管理文件系统名字空间的方法。该方法可以优化文件系统元数据访问性能。通过采集一个实际应用案例Apache服务的文件系统访问信息,揭示了应用程序与它所使用的文件之间的关联关系。采集的信息不但证明了应用语义的存在,给出了一种应用语义的提取方法,还说明了应用进程所面对的文件系统名字空间具有很大局部性。进而提出一种根据文件应用语义裁剪文件系统名字空间的方法。最后根据该方法设计并实现了一个原型文件系统。系统使用FUSE用户态文件系统框架提供文件系统功能接口,使用XML文件表示文件系统元数据,使用BDB XML数据库存储元数据XML文件,使用DOM树表示应用进程所面对的文件系统子名字空间。在原型文件系统中应用程序运行时所看到的是一个可确保其应用正常运行,但是大大简化了的文件系统子名字空间,这既改善了文件系统的元数据访问性能还提高了对文件语义预测的准确性。通过对比测试发现,同样在FUSE框架之中,相对于未对名字空间进行管理的文件系统,在使用了应用语义对名字空间进行管理的文件系统中一些应用程度的运行性能平均得到30%～50%的提升。更多还原

【Abstract】 Personal data is growing at ever increasing rates, fueled by a growing market for personal computing solutions and dramatic growth of available storage space on these platforms. File system as the base data management facility in operation systems for personal computer becomes more and more complex. File system metadata accessing performance is reduced due to the fairly complicated file system namespace.Complex namespace handling is important to file system performance. This work introduces a method for managing file system namespaces in an application-oriented way.We first use a popular application Apache Web Service as the case for tracing the relationship between applications and their files. The result of tracing file system shows that application oriented semantics actually exists, how to extract the application oriented semantics and proves that most of the applications only focus on a small part of the whole file system namespace. Then we make use of these application semantics to effectively simplify the namespace that applications must face, thus improve both the metadata accessing performance and file semantic prediction accuracy. We build a prototype file system for evaluating our method, analysis the potential problems and demonstrate the power and flexibility through a series of case studies, with a focus on how to manage file system namespace. In our prototype file system, we use FUSE which is a file system framework in userspace to provide file system function interface, use XML to represent file system metadata, use BDB XML database to store metadata XML files and use XML DOM tree to represent file system sub-namespace for a certain type of process.Our performance evaluation shows the overall gain is up to 30%~50% towards“raw”namespace which also using FUSE framework to some applications.更多还原