【作者】 王皓；

【导师】 李小勇；

【作者基本信息】 上海交通大学 ， 计算机技术， 2011， 硕士

【摘要】 随着信息技术的飞速发展,传统的存储系统已经难以满足现代社会的信息存储需求,很多新型应用需要使用大规模、高性能、高可靠的海量存储系统。近年来,学术界和企业界将海量存储系统技术作为研究的一个热点问题,并随之涌现出一批新型的海量存储系统。在海量存储系统中,客户端的设计与实现方式对于系统的性能等有重要影响。BlueOcean是一款大容量、高性能、高可靠、易管理的海量存储系统。它采用面向对象的存储技术,系统由管理节点、数据节点和客户端组成。管理节点负责保存和管理元数据,将元数据保存在内存以提升性能;数据节点负责保存数据,使用多份副本技术保证数据的可靠性;客户端对外提供遵守Posix语义的访问接口,可以支持绝大多数应用程序的透明访问。本文以BlueOcean海量存储系统为基础,研究了海量存储系统客户端设计中的关键技术问题,设计并实现了一款高性能的,通用的客户端程序。论文的主要工作如下:(1)对现有的海量存储系统进行了研究,分析了它们的系统架构和关键技术问题。(2)比较了几种客户端架构方案,设计了系统的总体架构以及通信交互协议等。(3)设计并实现了客户端的系统接口,描述了元数据操作接口和数据操作接口的工作流程,提出使用I/O流水线机制、预取、延迟写等技术提升读写性能。(4)分析了分布式系统中缓存设计的关键问题,设计并实现了客户端缓存机制来提升系统性能。(5)对BlueOcean海量存储系统的性能进行了测试,并且测试了缓存机制对于系统性能的影响。论文设计的客户端程序能够支持绝大多数应用程序对BlueOcean海量存储系统的透明访问,具有较好的元数据操作性能和数据读写性能,本论文的研究成果对于其它的海量存储系统客户端设计也具有较好的借鉴价值。更多还原

【Abstract】 With the rapid development of the information technology,the traditional storage system can’t satisfy the storage requirements in modern society,many new applications need the mass storage system with large scale,high performance and high reliability.In recent years,enterprises and academic community have treated mass storage technology as an hot point and developed several new mass storage systems.In mass storage system,the design and implementation of client program have important influence on system’s performance and generality.BlueOcean is a mass storage system with large scale,high performance, high reliability and easy management.BlueOcean is an Object-based storage system and is consists of management node,data node and client.Management node is responsible for storing meta data,it places all meta data in memory to improve performance.Data node is responsible for storing data,it uses replicas technology to ensure the reliability.Client provides the interfaces which obey the Posix standard,and it can support most application programs running on it.In this thesis,we made research on the critical technology of the mass storage system.We designed and implemented a high performance and generality client program based on BlueOcean.The main work in this thesis includes:(1)Analyse the current mass storage system,introduce their architecture and the critical technology.(2)Compare several client architectures,choose the File System in User Space to develop,design the system’s Framework and communication protocol.(3)Design and implement the client’s system interfaces,describe the working procedure of meta data operation and data operation,use the I/O pipeline mechanism, prefetch and delay write to improve read/write performance.(4)Introduce the critical technology of cache mechanism in distributed system,design and implement the cache mechanism to improve the system performance.(5)Test the BlueOcean mass storage system’s performance,and test the cache mechanism’s influence on performance.The Client software can support most application programs running on BlueOcean system.It has good meta data operation performance and data operation performance.The research is also valuable for designing client of other mass storage system.更多还原