【作者】 刘圆圆；

【导师】 魏晓辉；

【作者基本信息】 吉林大学 ， 计算机系统结构， 2016， 硕士

【摘要】 随着云计算的快速发展及其在各领域的广泛应用,作为云计算的基础支撑,云存储成为该领域的一个研究热点。云存储技术负责将数据资源存储到云端的数据中心,并为用户提供访问服务功能。云存储系统具有可用性、可靠性、扩展性以及快速的数据访问等特点,而系统的这些特征都与数据副本的管理密切相关。现有的成果多从不同目标出发,例如考虑系统能耗或用户的服务质量等,研究了云存储环境下的数据布局优化方法。为满足用户需求,大多数的云服务选择运行在多个地理上分布(Geo-distributed)的数据中心。在geo-distributed云存储服务中,用户将数据备份到云存储端并且有选择性地与其他用户共享,数据资源的存储是主要的成本开销,而服务的最终目标是为用户提供较好的用户体验。考虑到用户的区域性分布及用户量的爆炸性增长,将数据备份到多个云端来减少访问数据时的传输时间,使用户获得更好的用户体验,是更有效的方法。然而,数据副本备份又会导致更高的存储花费,因此,在数据的存储开销和用户体验之间进行较好的权衡是云存储服务所面临的挑战。我们提出的数据副本放置策略,需要既能保证较少的存储开销的同时,又可以兼顾用户体验性能。近年来,移动互联网的便利性促进了移动用户数量的快速增加。更多用户经常会利用移动终端,随时随地备份数据到云存储中心,移动云存储的研究也逐渐被大家关注。而移动性也带来了新的挑战,例如,移动云存储服务的推广促进了用户之间对数据的相互访问,用户位置不固定影响了用户体验等。因此,系统放置数据时也需考虑云存储服务的移动性。用户作为云存储服务主体,要求系统在存储副本时还需考虑用户的优先级和活跃度等行为特征。如何应对移动互联网,考虑用户特性对数据副本放置的影响是亟待解决的问题。本文主要针对云存储服务的地理分布特性,结合用户行为特征,根据上述问题,做了以下工作:(1)在本文中,我们对geo-distributed云存储服务中的数据副本放置问题进行建模,主要从云存储服务提供商的角度出发,综合考虑了数据的存储开销和用户体验,而本文的用户体验性能是指用户访问数据的传输延迟。本文不仅考虑了云存储中心的地域分布性,而且详细分析了用户的访问特征。结合实际应用情况,用户会访问其他用户共享的文件,用户之间以文件为纽带建立了联系,基于此重点考虑了用户行为对数据存储的影响。(2)通过在存储开销和用户体验之间权衡,提出了数据放置的优化研究策略,设计了数据副本的调整算法(RPA),结合用户和云存储中心的特性,可将该算法用于定期优化系统的整体存储情况。(3)当用户要放置副本到云存储中心时,本文考虑了用户和数据中心,用户之间的通信对副本存储的影响,分析了geo-distributed云存储服务的社会感知特征。提出了基于社会感知的数据副本放置算法(DDRP),算法根据当前环境下的用户状态,及用户间的关系等信息动态地存储用户数据副本。根据该算法系统可以选择较优的数据中心集来放置数据副本。实验结果显示,算法RPA在不影响用户体验的同时,调整副本后减少存储开销约45%。与其他副本放置方法相比,本文提出的DDRP算法能够在降低存储开销的同时又保证较好的用户体验性能。与其他方法相比数据副本存储的整体开销减少约40%,验证了算法的有效性。更多还原

【Abstract】 With the rapid development and the widespread use of cloud computing in various applications, as a basis support for cloud computing, the cloud storage has become a research hotspot in this field. The cloud storage technique is responsible for putting the stored resources in the datacenters of clouds for users to access. Cloud storage system has the characteristics of reliability, scalability, and fast data access, and these features of systems are closely related with the management of data replicas. Existing results studied on the optimization methods about data layout in cloud computing environments mostly from different objectives, such as considering energy consumption of systems and quality of service of users.To meet the users’ requirement, most cloud services choose to run on the multiple geographically distributed(geo-distributed) datacenters. In geo-distributed cloud storage service, users choose these clouds to back up their data and selectively share with others. And the data storage is the major monetary expenditure while the ultimate goal is to provide users with a better user experience. Considering regional distribution of the locations of users and explosive growth of users, it’s more efficient for users to replicate data to many distributed clouds to transmission time of accessing data to obtain a better user experience. However, data replicas also brings a problem of higher storage cost. Thus, making a better trade-off between storage cost and user experience is a challenge faced by cloud storage service. We need to propose effective data replica placement strategies, which need to guarantee less cost overheads while taking user experience into account.In recent years, the convenience of the mobile Internet promotes the rapid increase in the number of mobile users. More users will usually backed up their data to the cloud storage center, anytime and anywhere. Research on mobile cloud storage has gradually been concerned. However, mobility also brings new challenges, for instance, the use of mobile cloud storage has promoted the interactions between users for data, flexibility of users’ locations affects the user experience etc. Thus, the system need consider the mobility of cloud storage service when placing data. Because users can be seen as a main part of cloud storage service, which makes it necessary for the system to consider user access features such as users priority and activeness when storing data replicas. How to cope with mobile Internet, and consider the effect of user characteristics on replica placement is a problem to be solved.This article focuses on the geo-distributed cloud storage services. According to the abovementioned questions and based on user behavior characteristics, the following work is as follows:(1). In this paper, we modeled the optimization of data replica placement in geo-distributed cloud storage service mainly from the perspective of cloud service providers. We both take the data storage cost and user experience into account. This paper not only considers the geographical distribution of cloud storage centers, but analyzes the access features of users in detail. Combined with practical application, user files may be shared with other users that can be seen as a link to establish contacts between users. Based on these, the impact of user behavior on the storage of data has been mainly described.(2). We modeled the problem to do a tradeoff between storage and user experience, formulated the optimization objective function of data placement, and then designed a multiple replica adjustment algorithm. Based on the characteristics of users and cloud storage centers, the algorithm can be used to regularly optimize the overall data storage of systems.(3). When placing replicas to clouds, we considered the communications between users and clouds, between users, and analyzed some features of social-aware cloud storage services. We presented a replica placement algorithm(DDRP) for social-aware service, which can store data replica dynamically based on the current information about users’ state and the relationship between users. Leveraging the algorithm, a better replica placement scheme to choose which cloud to place data can be received.Simulation results show that, the RPA algorithm can reduce storage cost about 45% after adjusting replica while not affecting user experience. Compared with other replicas placement method, the proposed DDRP algorithm in our paper can ensure a better user experience performance while reducing storage costs. The overall cost of storing data replica is 40% less than other schemes, which demonstrates its effectiveness.更多还原