云环境下基于用户请求响应时间的副本管理策略研究

发布时间：2018-06-19 14:22

本文选题：云存储 + 副本管理　；参考：《华中科技大学》2013年硕士论文

【摘要】：随着信息技术的快速发展，数据规模呈指数级增长，为了存储和管理海量数据，云存储系统应运而生。云存储系统将分布广泛的存储资源进行有效的组织，以较为低廉的成本，向用户提供大容量、高性能的存储服务。副本技术是云存储系统的重要组成部分，是云存储系统实现可靠性、并发性和扩展性的基础。目前的副本管理策略研究中，对于单个用户请求的响应时间没有引起足够重视，造成许多请求长时间处于等待状态，降低了系统的性能。基于用户请求响应时间的副本管理策略以减少用户请求的响应时间为目标，以云存储网络的拓扑结构为基础，在副本放置、副本选择、动态副本创建和副本删除等方面进行了深入研究。副本放置过程中考虑响应时间和节点的存储能力，保证其他节点都能在规定时间内访问到该数据文件。副本选择过程中综合考虑副本节点的存储能力以及副本节点与用户节点的距离，选择服务能力最强的副本节点响应用户请求。动态副本创建根据用户访问频率，动态调整文件的扫描周期，对系统中的文件实行差异化管理；另外，根据用户访问频率判断系统中是否出现“热点数据”，并为“热点数据”创建新的副本。副本删除以用户访问频率为依据，在保证数据可靠性的前提下，删除系统中长时间没有用户访问的副本。通过模拟器OptorSim进行模拟实验，将基于用户请求响应时间的副本管理策略与OptorSim内置的副本管理策略进行比较，试验结果表明，基于用户请求响应时间的副本管理策略的平均响应时间比OptorSim降低了32%，，网络利用率比OptorSim降低了48%，系统中的副本总数减少了16%。
[Abstract]:With the rapid development of information technology, the scale of data increases exponentially. In order to store and manage mass data, cloud storage system emerges as the times require. Cloud storage system will be widely distributed storage resources for effective organization, at a relatively low cost, to provide users with large capacity, high performance storage services. Replica technology is an important part of cloud storage system, and it is the basis of reliability, concurrency and expansibility of cloud storage system. In the current replica management policy research, the response time for a single user's request is not paid enough attention to, resulting in a lot of requests waiting for a long time, which reduces the performance of the system. The replica management strategy based on user request response time aims to reduce the response time of user request, and based on the topology of cloud storage network, the replica is placed in the replica, and the replica is selected. Dynamic replica creation and replica deletion are studied in depth. The response time and the storage capacity of the nodes are considered in the process of replica placement to ensure that other nodes can access the data file within the specified time. In the process of replica selection, the storage capacity of replica node and the distance between replica node and user node are considered synthetically, and the replica node with the strongest service ability is selected to respond to user request. Dynamic replica is created according to the frequency of user access, dynamically adjusts the scanning period of files, and implements differential management of files in the system; in addition, according to the frequency of user access, it determines whether there are "hot data" in the system. And create a new copy of Hot data. The replica deletion is based on the frequency of user access. Under the premise of ensuring the reliability of the data, the replica which has not been accessed by the user for a long time in the system is deleted. The replica management strategy based on user request response time is compared with the replica management policy built in by OptorSim. The experimental results show that, The average response time of the replica management policy based on user request response time is 32 times lower than OptorSim, the network utilization ratio is lower than OptorSim by 48 percent, and the total number of replicas in the system is reduced by 16 percent.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【参考文献】