云环境下副本策略研究

发布时间：2018-06-22 06:56

本文选题：云存储 + 数据中心　；参考：《吉林大学》2015年硕士论文

【摘要】：随着信息技术的高速发展，计算机存储数据的规模已出现爆炸性增长的趋势，云存储凭借其扩展性强、容错性好、性价比高等优势，成为了业界研究关注的热点。云存储系统通常由多个数据中心的存储资源构成，且作为分布式存储平台，其底层硬件设施由大规模廉价服务器构成，如何保证系统的容错性，进而提高数据的可靠性是云存储中一个很重要的研究问题。副本技术是当前提高系统容错性能的主流技术，通过为一份数据创建多份副本的方式来解决单点失效时带来的数据失效问题。同时，副本技术有助于降低访问延迟，提高网络带宽利用率，也有利于提高数据的可用性。当前关于副本技术的研究可以分为如下两类：静态副本策略和动态副本策略。目前大部分系统均采用静态副本策略，即在数据进入前已经确定副本的数量和位置，这种方式虽然简单，但缺少灵活性，且没有考虑用户需求、存储节点自身的性质以及周围环境的变化等因素的影响，在一定程度上会造成存储资源的不合理利用。为了更加合理地利用云中的存储资源，丰富副本技术的研究，本文对云存储中的数据备份策略进行了研究，提出了一种基于关联失效的数据动态备份策略。本文的研究内容具体包含以下几部分：(1)对传统的以交换机为中心的分层网络拓扑结构和微软研究院提出的VL2树型结构进行了深入分析，基于分析结果构建了数据中心网络结构模型和功能结构模型；(2)综合距离因素和网络、制冷、电力、自然灾害等环境因素分析了存储节点间关联性，基于分析结果构建了节点间的关联失效模型，结合节点自身的属性和节点间的级联响应构建了单点失效率计算模型，并基于节点间关联失效模型和单点失效率计算模型给出了衡量任一数据在不同存储方案下可靠性的计算方法；(3)不同用户对于云计算中存储数据的可靠性需求存在差别，在充分考虑这一前提的情况下提出了一种基于关联失效的数据动态备份策略，该策略综合分析地理、环境等因素对于数据中心中节点间关联性的影响，在保证用户可靠性需求的前提下尽可能地降低数据的副本数量，进而提高存储资源的利用率。实验表明，与Amazon的静态副本备份策略相比，，本文提出的方法充分考虑了用户需求和节点间的关联性，更符合实际的存储情况，在满足用户对于数据可靠性需求的同时，一定程度上降低了该系统下数据的存储空间，有效地提高了存储资源的利用率。
[Abstract]:With the rapid development of information technology, the scale of computer storage data has been explosive growth trend, cloud storage with its strong expansibility, good fault tolerance, high cost performance and other advantages, has become the focus of attention in the industry. Cloud storage system is usually composed of storage resources of multiple data centers, and as a distributed storage platform, its underlying hardware facilities are composed of large scale cheap servers, how to ensure the fault tolerance of the system. Improving the reliability of data is an important research problem in cloud storage. Replica technology is the main technology to improve the fault-tolerant performance of the system at present. It solves the problem of data failure caused by single point failure by creating multiple copies for one data. At the same time, replica technology can reduce access delay, improve network bandwidth utilization and improve data availability. The current research on replica technology can be divided into two categories: static replica strategy and dynamic replica strategy. At present, most systems adopt a static replica strategy, that is, the number and location of replicas are determined before the data enters, which is simple, but inflexible, and does not take into account the needs of users. The influence of the properties of storage nodes and the changes of surrounding environment will lead to unreasonable utilization of storage resources to a certain extent. In order to make more rational use of storage resources in the cloud and enrich the research of replica technology, this paper studies the data backup strategy in cloud storage, and proposes a dynamic backup strategy based on association failure. The research contents of this paper include the following parts: (1) the traditional hierarchical network topology with switch as the center and the VL2 tree structure proposed by Microsoft Research Institute are deeply analyzed. Based on the analysis results, the network structure model and the functional structure model of the data center are constructed. (2) the relationship between storage nodes is analyzed based on the environmental factors such as distance factors and network, refrigeration, electricity, natural disasters, etc. Based on the analysis results, an association failure model between nodes is constructed, and a single point failure rate calculation model is constructed by combining the attributes of nodes and the cascade response between nodes. And based on the inter-node associated failure model and single-point failure rate calculation model, a calculation method to measure the reliability of any data under different storage schemes is given. (3) different users have different reliability requirements for data stored in cloud computing. In this paper, a dynamic data backup strategy based on association failure is proposed, which comprehensively analyzes the influence of geography, environment and other factors on the relationship between nodes in the data center. In order to improve the utilization of storage resources, the number of replicas of data can be reduced as much as possible on the premise of ensuring the reliability of users. The experiments show that compared with the static copy backup strategy of Amazon, the method proposed in this paper fully considers the relationship between user requirements and nodes, and is more in line with the actual storage situation, and meets the data reliability requirements of users at the same time. To some extent, the data storage space is reduced, and the utilization rate of storage resources is improved effectively.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP333

【参考文献】