基于网络编码的分布式数据存储与故障修复策略

发布时间：2018-04-28 09:21

本文选题：网络编码 + 分布式存储　；参考：《浙江工业大学》2013年硕士论文

【摘要】：如今,人类社会已经步入了信息化高速发展的时期,信息在社会生活中起举足轻重的地位。与此同时,海量信息数据的存储也越来越被人们所重视。尤其是近几年高清视频的广泛应用,大数据时代的来临,传统的集中式数据存储系统呈现出越来越多的弊端,比如安全性不高,可靠性也没有保证,可扩展性不强等。分布式数据存储系统则弥补了这方面的缺陷,它使得海量数据能以分布的形式存储,满足了大规模存储的需求,而且具有更高的稳定性。但是分布式数据存储会面临存储节点故障的问题,因而如何修复故障节点保证系统的稳定运行成为分布式数据存储系统的重要课题。当前,修复故障节点的指标主要有节点存储容量,修复带宽,修复开销,重建时间等。如何在修复故障节点的同时使得相关指标尽可能地小成为了当前分布式数据存储的研究热点。本文研究了基于网络编码的分布式数据存储及故障修复策略,主要的工作和贡献如下： (1)本文将分布式数据存储系统的故障修复问题抽象化为基于网络编码的信息流图,以便于用图论的相关理论来简化分析数据传输过程。 (2)针对目前的故障修复策略没有对重建时间进行优化,本文从理论上分析了重建时间的下界,然后提出了最小化重建时间故障修复策略MTR (Minimum-Time Repair),该修复策略达到了重建时间的下界值,而且,在此基础上将修复带宽降到了最低值。分析数据表明,与现有的节点修复策略相比,MTR修复策略能大大降低重建时间,使得故障节点快速修复,这对那些需要快速修复的分布式数据存储系统有着非常重要的意义。 (3)考虑到实际应用中,重建时间受约束这个因素,本文提出了重建时间约束的带宽最优故障修复策略,建立了该策略的优化模型并显式给出最优解。该策略能在给定的重建时间阈内修复故障节点,且在该时间点上的修复带宽是最小的。同时,该策略具有较好的灵活性,在任何重建时间阈值(大于重建时间下界)内,都存在使修复带宽最小的故障修复策略。
[Abstract]:Nowadays, human society has stepped into the period of rapid development of information, and information plays an important role in social life. At the same time, people pay more and more attention to the storage of massive information data. Especially in recent years, with the wide application of high-definition video and the advent of big data era, the traditional centralized data storage system presents more and more disadvantages, such as low security, no guarantee of reliability, low scalability and so on. Distributed data storage system makes up for the defects in this aspect. It enables the mass data to be stored in the form of distribution, which meets the needs of large-scale storage and has higher stability. However, distributed data storage will face the problem of storage node failure, so how to repair the fault node to ensure the stable operation of the system becomes an important issue of distributed data storage system. At present, the main indexes of repairing fault nodes are node storage capacity, repair bandwidth, repair overhead, reconstruction time and so on. How to repair the fault nodes and make the relevant indexes as small as possible has become the research hotspot of distributed data storage. This paper studies the distributed data storage and fault repair strategy based on network coding. The main work and contributions are as follows: In this paper, the fault repair problem of distributed data storage system is abstracted into an information flow graph based on network coding, so as to simplify the analysis of data transmission by using the relevant theory of graph theory. 2) since the current fault repair strategy does not optimize the reconstruction time, this paper theoretically analyzes the lower bound of the reconstruction time. Then, the minimum reconstruction time fault repair strategy MTR / Minimum-Time repair strategy is proposed, which reaches the lower bound of the reconstruction time, and reduces the repair bandwidth to the lowest value on this basis. The analysis results show that the MTR repair strategy can greatly reduce the reconstruction time and make the fault nodes repaired quickly compared with the existing node repair strategy which is of great significance to those distributed data storage systems which need to be repaired quickly. Considering the fact that the reconstruction time is constrained in practical applications, this paper proposes an optimal fault repair strategy with reconstruction time constraints, and establishes an optimal model of the strategy and gives the optimal solution explicitly. The proposed strategy can repair the fault nodes within the threshold at a given reconstruction time, and the repair bandwidth at this time point is minimal. At the same time, the strategy has good flexibility, in any reconstruction time threshold (larger than the reconstruction time lower bound), there is a fault repair strategy to minimize the repair bandwidth.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【相似文献】