【作者】 杨伟；

【导师】 齐小刚；

【作者基本信息】 西安电子科技大学 ， 应用数学， 2018， 硕士

【摘要】 随着信息社会的发展,互联网发展迎来爆炸时代,数据量增长加快。传统的集中式数据存储技术由于受到设备能力的限制,已经无法适用于目前的情况。分布式存储系统以其良好的可扩展性,健壮性和高效性超越了传统的集中式存储技术。目前的数据存储多考虑存取效率、存储花费,对数据抗毁性研究甚少,如何使得分布式数据存储系统具备优良的鲁棒性是当今面临的挑战之一。本文首先对分布式数据存储系统进行介绍,并分别介绍其两个具体表现:云存储系统和对等(P2P)存储系统。接着介绍了分布式数据存储系统数据存储方法和一致性维护方法,分别对其相关研究进行分类,提出分类标准并对相关研究方法详细剖析,分析其适用场景和面临的挑战。针对分布式数据存储系统数据存储和数据抗毁性研究,我们做了以下两部分具体实现:针对分布式数据存储系统数据存储方法方面,考虑到数据泄露问题已经是目前数据存储系统面临的最大挑战之一,本文提出一个考虑数据存取效率和安全性的数据存储方法。提出安全距离的概念不仅保证了数据的安全性,而且考虑到了不同用户群体对于数据安全的侧重性。接着为了最小化数据存取时间,本文提出一个数据存取节点选择算法,在保证数据安全性的同时最小化数据存取时间。仿真证明相比目前的几种数据存储算法,在保证数据安全性的同时,本文提出的算法在数据存取时间方面小于目前的数据存储算法。针对分布式数据存储系统数据一致性维护方面,为了解决由于网络动荡和网络打击引起的丢包问题,本文提出一个可调节且可靠的一致性维护方法(ARCOM)。ARCOM通过构造更新分发环(mUD-Ring)来传播更新。mUD-Ring运用以下四种方法提高一致性维护效率:多次询问方案避免丢包产生、多点传输可以达到更高的更新传播效率、发送更新前询问可以减少数据冗余、在节点中开辟缓存区能够避免更新丢失。考虑到节点打击和链路损坏,本文提出一个低花费的恢复策略并且对其作了理论分析。Omnet++仿真表明ARCOM与现有一致性维护方法相比有着高的更新传播效率、低的丢包率(近乎为0,其他方法在15%-80%之间)、低的数据冗余。更多还原

【Abstract】 With the development of the information society,the development of the Internet ushered in the age of explosion,and the amount of data has accelerated.Traditional centralized data storage technology has been unable to adapt to the current situation due to limited equipment capabilities.Distributed storage systems surpass traditional centralized storage technologies with their excellent scalability,robustness,and efficiency.The current data storage methods consider the access efficiency and storage cost,and there is little research on data invulnerability.How to make the distributed data storage system have excellent robustness is one of the challenges facing today.This article first introduces the distributed data storage system and introduces two specific performances: cloud storage system and peer-to-peer(P2P)storage system.Then introduces the data storage method and consistency maintenance method of distributed data storage system,classifies related research,puts forward the classification standard and analyzes the relevant research methods in detail,and analyzes its application scenarios and challenges.For the data storage method of distributed data storage system,taking into account the data leakage problem is one of the biggest challenges facing the current data storage system,this paper proposes a data storage method that considers data access efficiency and security.The concept of safety distance proposed not only ensures the security of data,but also takes into account the different requirements of different user groups for data security.To minimize data access time,a node selection algorithm is proposed in this paper to minimize data access time while ensuring data security.The simulation proves that in terms of data access time,compared with the current data storage algorithms,the proposed algorithm is less than the current algorithm in terms of data access time while ensuring data security.For the data consistency maintenance of distributed data storage systems,in order to solve the problem of packet loss caused by network turbulence and network attacks,this paper proposes an adjustable and reliable consistency maintenance method(ARCOM).ARCOM propagates updates by constructing updates distribution ring(m UD-Ring).m UD-Ring uses the following four methods to enhance the consistency maintenance efficiency,including multiple query schemes to avoid packet loss,multi-point transmission that can achieve higher transmission efficiency,inquiring before distributing updates with the aim at reducing data redundancy,opening a cache in a node,which can avoid the loss of updates.Considering the failure of nodes and links,the paper presents a low-cost recovery strategy and makes a theoretical analysis.Simulation results through Omnet++ platform show that the ARCOM has a higher transmission efficiency,lower packet loss rate(close to 0,the other methods are between 15% and 80%)and lower data redundancy compared with the several current popular consistency maintenance schemes.更多还原