基于DHT的存储系统中纠删码技术研究

发布时间：2018-01-03 08:46

本文关键词：基于DHT的存储系统中纠删码技术研究　出处：《西安电子科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：基于分布式哈希表（DHT）的云存储系统有着良好的扩展性和快速的数据存取能力。基于DHT的数据存储系统具有动态性和异构性，导致数据随时可能丢失。因此，如何保证DHT系统的高可用性成为研究DHT存储系统的关键问题。本文研究了基于DHT的云存储系统的冗余机制和基于柯西矩阵的纠删码技术，并改进了柯西编码的编码运算和数据更新策略。本文的主要工作概括如下： 1.概述了分布式哈希表的设计原理和几种常见DHT协议的基本原理，分析了传统的副本和纠删码两种数据冗余机制，阐述了运用在DHT存储系统中的RS纠删码的原理。 2.分析了基于柯西矩阵的RS纠删码的原理，根据柯西编码过程中有限域运算的特点给出了一种减少运算过程中异或操作个数的算法。实验结果表明通过该方法有效地减少了运算过程中异或操作的个数，从而相对传统RS码而言有着更优的编码性能。 3.仔细分析了现有的两种数据更新策略，并给出了一种新的数据更新策略。通过将数据更新过程中的运算分布到多个节点来减少更新过程中运算时间，，有效地提高了数据更新的效率。仿真结果表明本文所给出的数据更新策略比传统的更新策略有着更高的效率。
[Abstract]:The cloud storage system based on distributed hash table (DHT) has good expansibility and fast data access ability. The data storage system based on DHT is dynamic and heterogeneous. Therefore, how to ensure the high availability of DHT system becomes a key problem in the study of DHT storage system. This paper studies the redundancy mechanism of cloud storage system based on DHT and erasure code technology based on Cauchy matrix, and improves the coding operation and data updating strategy of Cauchy coding. The main work of this paper is summarized as follows:. 1. The design principle of distributed hash table and the basic principles of several common DHT protocols are summarized, and two kinds of data redundancy mechanisms, traditional replica and erasure code, are analyzed. The principle of RS erasure code used in DHT storage system is described. 2. The principle of RS erasure code based on Cauchy matrix is analyzed. According to the characteristics of the finite field operation in the Cauchy coding process, an algorithm is proposed to reduce the number of XOR operations in the operation process. The experimental results show that this method can effectively reduce the number of XOR operations in the operation process. Compared with the traditional RS code, it has better coding performance. 3. Two existing data update strategies are analyzed carefully, and a new data update strategy is proposed. The operation time is reduced by distributing the operation in the process of data updating to multiple nodes. The simulation results show that the data update strategy presented in this paper is more efficient than the traditional updating strategy.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333;TN911.2

【参考文献】