基于对象的pNFS分级存储设计和实现

发布时间：2018-04-21 14:37

  本文选题：分布式文件系统 + 并行网络文件系统　； 参考：《国防科学技术大学》2012年硕士论文

【摘要】：随着经济社会的发展，传统高性能存储系统如存储局域网（SAN）和网络附加存储（NAS）在性能和扩展性上难以满足海量数据的处理要求。基于对象存储设备（OSD）的并行网络文件系统（pNFS）通过元数据和数据的分离，有效弥补了网络文件系统（NFS）性能和扩展性上的不足。 用户对存储系统的需求也从单纯的高性能发展到综合考虑存储系统的性能价格比，对扩展性和可靠性的需求也与日俱增。另一方面，随着业务发展，，用户的存储系统往往同时存在不同性能级别的存储设备，新型的高速存储设备应该用于尽可能地提高系统整体性能。分级存储技术正是这些问题的一种有效解决方式。 基于此现状，本文在深入分析当前主流网络存储技术和现有典型分布式存储系统的基础上，采用pNFS协议，提出了一种融合分级存储技术和对象存储技术的分布式文件系统模型t-pNFS。该模型充分利用对象存储设备的灵活性，具有高性能、高可靠、可横向扩展等特点，适用于对读性能和吞吐率要求较高的领域。该模型提出了多种性能提升手段，主要包括：适用于缓存式分级的热点数据识别算法和缓存策略，缓存回收策略等。 文章的最后，通过对比测试，验证了t-pNFS模型的有效性。
[Abstract]:With the development of economy and society, traditional high performance storage systems such as storage LAN (San) and network attached storage (NASS) can not meet the requirements of mass data processing in terms of performance and scalability. Parallel network file system (pNFS) based on object storage device (OSD) can effectively compensate for the performance and expansibility of network file system (NFS) by separating metadata from data. Users' demand for storage system has also developed from pure high performance to comprehensive consideration of the performance / price ratio of storage system, and the demand for scalability and reliability is also increasing day by day. On the other hand, with the development of business, user storage systems often have different performance levels of storage devices at the same time, new high-speed storage devices should be used to improve the overall performance of the system as much as possible. Hierarchical storage technology is an effective solution to these problems. Based on this situation, this paper presents a distributed file system model t-pNFS, which combines hierarchical storage technology with object storage technology, based on in-depth analysis of current mainstream network storage technology and existing typical distributed storage systems, using pNFS protocol. The model makes full use of the flexibility of object storage devices, and has the characteristics of high performance, high reliability and lateral expansion. It is suitable for areas with high read performance and throughput. This model proposes a variety of performance enhancement methods, including: the hot data recognition algorithm and cache strategy for cache classification, cache recovery strategy and so on. At the end of the paper, the validity of t-pNFS model is verified by contrast test.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP333

【参考文献】

中国期刊全文数据库 前5条

1 杨德志,黄华,张建刚,许鲁;大容量、高性能、高扩展能力的蓝鲸分布式文件系统[J];计算机研究与发展;2005年06期

2 黄华;张建刚;许鲁;;蓝鲸分布式文件系统的客户端元数据缓存模型[J];计算机科学;2005年09期

3 孟晓p

本文编号：1782850