基于SSD和HDD的分级文件系统的设计与实现

发布时间：2018-05-24 07:14

本文选题：分级存储 + 文件系统　；参考：《北京邮电大学》2014年硕士论文

【摘要】：存储数据的飞速增长,要求企业的信息系统不仅要提供更大的存储容量：同时还要提供更高的存储速度。分级存储是当前存储系统中解决容量和速度问题的重要技术。面对海量数据的存储需求,分级存储技术针对数据特性设计相应的存储系统,对价值较高的数据和对系统性能影响较大的数据进行优化处理,以提供更好的性能。而固态盘(Solid State Drive, SSD)的广泛应用为设计现代分级存储系统带来了新的挑战。固态盘是一种基于闪存.的新型存储介质,提供了比传统磁盘(Hard Disk Drive, HDD)更高的I/O性能,同时固态盘自身也引入了一些新的问题,如寿命磨损问题和成本高昂问题。为了有效地将固态盘和传统磁盘整合起来,构建高效的分级存储系统,本文对传统存储技术、固态盘技术和Linux文件系统进行了研究,设计并开发了一个基于SSD和HDD的分级文件系统FMCFS。FMCFS文件系统实现了对固态盘空间和磁盘空间的高效管理,并能统计出用户对文件数据块的长期历史访问信息,识别出文件系统中的热点数据块,并在文件数据块级别对固态盘和磁盘中的数据进行自动的迁移和回迁。本文对设计分级文件系统的关键问题进行了研究,设计并实现了高效管理大规模元数据的策略、识别文件中热点数据块的算法,以及在内核中通过I/O重定向进行数据迁移和回迁的方法。本文的结果显示,分级文件系统FMCFS把高访问频率和高价值的文件数据块自动地迁移到固态盘上,可以充分利用固态盘的高I/O吞吐量特性和磁盘的大容量低价格特性,从而提供了一种性价比合理的高性能自动分级存储方案,并能有效减少固态盘的磨损问题对存储系统的影响。
[Abstract]:With the rapid growth of storage data, enterprise information systems should not only provide greater storage capacity, but also provide higher storage speed. Hierarchical storage is an important technology to solve the problem of capacity and speed in current storage system. In order to provide better performance, the hierarchical storage technology designs the corresponding storage system in view of the characteristics of the data, and optimizes the processing of the high-value data and the data that has a great impact on the performance of the system in order to provide better performance. The wide application of solid State Drive, SSD) brings new challenges to the design of modern hierarchical storage system. A solid-state disk is based on flash memory. The new storage media provides better I / O performance than the traditional hard Disk Drive, HDD), while the solid-state disk itself introduces some new problems, such as lifetime wear and high cost. In order to integrate the solid-state disk and the traditional disk effectively and build an efficient hierarchical storage system, the traditional storage technology, solid-state disk technology and Linux file system are studied in this paper. A hierarchical file system, FMCFS.FMCFS file system based on SSD and HDD, is designed and developed to efficiently manage the solid disk space and disk space. The hot data blocks in the file system are identified, and the data in the solid disk and disk are automatically migrated and returned at the file data block level. In this paper, the key problems of designing hierarchical file system are studied, and the strategy of efficiently managing large-scale metadata and the algorithm of identifying hot data blocks in files are designed and implemented. And the method of data migration and return through I / O redirect in kernel. The results of this paper show that the hierarchical file system FMCFS automatically migrates the high access frequency and high value file data blocks to the solid state disk, which can make full use of the high I / O throughput characteristics of the solid state disk and the large capacity and low price characteristics of the disk. Thus, a high performance automatic hierarchical storage scheme with reasonable performance and price ratio is provided, and the effect of solid state disk wear on storage system is effectively reduced.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP333

【参考文献】