云存储环境下数据副本技术研究
发布时间：2018-11-17 11:09
【摘要】：云存储旨在将企业和个人从繁重的数据存储工作中分离出来，以“按需付费”的形式向用户提供简单而可靠的存储服务，目前正在受到越来越多的关注。副本技术是保证云存储中数据安全和可靠的关键技术之一。云存储副本技术的研究，对云存储的发展和推广具有重要意义。 本文以开源的HDFS为平台，分析了云存储中副本管理的特点和目前存在的问题。针对云存储中用户的多变性，，本文给出一种基于文件热度的动态副本冗余度策略。该策略综合文件的历史访问信息，通过对不同周期的访问频率赋予不同的权重来计算得到文件的当前热度值。文件的副本冗余数以文件的热度值为依据，根据文件热度值的变化而动态变化。随后对副本数量的最大值和最小值进行了理论分析，提出了依据用户的可靠性要求来确定副本数量的最小值，依据系统允许的一致性维护开销确定副本数量最大值的方法。 针对云存储中节点的差异性，本文设计了基于TOPSIS的副本布局策略，将其运用在副本布局策略中以充分利用其在多目标决策上的优势。基于TOPSIS的副本布局策略使用HDFS机架感知的功能，对多个节点的多个属性进行综合评价，通过计算各个节点到最优解和最劣解的“距离”来选择部署副本的节点，从而达到副本布局的高效性。
[Abstract]:Cloud storage, which aims to separate enterprises and individuals from heavy data storage work and provide users with simple and reliable storage services in the form of "on-demand payment", is attracting more and more attention. Replica technology is one of the key technologies to ensure data security and reliability in cloud storage. The research of cloud storage replica technology is of great significance to the development and popularization of cloud storage. Based on open source HDFS, this paper analyzes the characteristics and problems of replica management in cloud storage. In view of the variability of users in cloud storage, this paper presents a dynamic replica redundancy strategy based on file heat. The strategy synthesizes the historical access information of the file and gives different weights to the access frequency of different periods to calculate the current heat value of the file. The redundant number of copies is based on the calorific value of the file and changes dynamically according to the change of the calorific value of the document. Then the maximum and minimum replicas are theoretically analyzed and a method is proposed to determine the minimum replica number according to the reliability requirements of the user and the maximum replica number according to the allowable consistency maintenance overhead of the system. Aiming at the differences of nodes in cloud storage, this paper designs a replica layout strategy based on TOPSIS, and applies it to the replica layout strategy to make full use of its advantages in multi-objective decision making. Based on the replica layout strategy of TOPSIS, using the function of HDFS rack awareness, the multiple attributes of multiple nodes are comprehensively evaluated, and the nodes that deploy replica are selected by calculating the "distance" from each node to the optimal solution and the worst solution. In order to achieve the high efficiency of the copy layout.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333
本文编号：2337571
[Abstract]:Cloud storage, which aims to separate enterprises and individuals from heavy data storage work and provide users with simple and reliable storage services in the form of "on-demand payment", is attracting more and more attention. Replica technology is one of the key technologies to ensure data security and reliability in cloud storage. The research of cloud storage replica technology is of great significance to the development and popularization of cloud storage. Based on open source HDFS, this paper analyzes the characteristics and problems of replica management in cloud storage. In view of the variability of users in cloud storage, this paper presents a dynamic replica redundancy strategy based on file heat. The strategy synthesizes the historical access information of the file and gives different weights to the access frequency of different periods to calculate the current heat value of the file. The redundant number of copies is based on the calorific value of the file and changes dynamically according to the change of the calorific value of the document. Then the maximum and minimum replicas are theoretically analyzed and a method is proposed to determine the minimum replica number according to the reliability requirements of the user and the maximum replica number according to the allowable consistency maintenance overhead of the system. Aiming at the differences of nodes in cloud storage, this paper designs a replica layout strategy based on TOPSIS, and applies it to the replica layout strategy to make full use of its advantages in multi-objective decision making. Based on the replica layout strategy of TOPSIS, using the function of HDFS rack awareness, the multiple attributes of multiple nodes are comprehensively evaluated, and the nodes that deploy replica are selected by calculating the "distance" from each node to the optimal solution and the worst solution. In order to achieve the high efficiency of the copy layout.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333
【参考文献】
相关期刊论文 前6条
1 郭东;杜勇;胡亮;;基于HDFS的云数据备份系统[J];吉林大学学报(理学版);2012年01期
2 周旭;卢显良;侯孟书;詹川;;频率自适应的动态副本管理机制[J];计算机科学;2005年02期
3 侯孟书;王晓斌;卢显良;任立勇;;一种新的动态副本管理机制[J];计算机科学;2006年09期
4 董继光;陈卫卫;田浪军;吴海佳;;大规模云存储系统副本布局研究[J];计算机应用;2012年03期
5 周可;王桦;李春花;;云存储技术及其应用[J];中兴通讯技术;2010年04期
6 薛一波;易成岐;;云存储(2)[J];中兴通讯技术;2012年02期
相关博士学位论文 前1条
1 谭支鹏;对象存储系统副本管理研究[D];华中科技大学;2008年
相关硕士学位论文 前2条
1 王彩亮;云存储环境下数据副本管理策略研究[D];云南大学;2011年
2 陈虎;基于HDFS的云存储平台的优化与实现[D];华南理工大学;2012年
本文编号：2337571
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2337571.html
