【作者】 张磊；

【导师】 李学俊；

【作者基本信息】 安徽大学 ， 软件工程， 2018， 硕士

【摘要】 云计算是一种以国际网络相关服务为基础的资源共享计算模式。云计算平台是以数据存储、数据处理为主的计算型平台并且在共享海量分布式资源的同时向用户提供按需使用的网络存储计算资源服务。在线社交网络平台是一种由计算机存储和处理的服务平台,在社交网络中服务提供商需要根据用户之间的联系和位置为每个用户提供合理的资源分配并且随着云环境中社交网络平台的不断发展,使得对云环境中资源调度分配和执行的QoS(Quality of Service)目标的要求不断提升。因此,如何使得云环境中在线社交网络的资源分配方案更加合理是一个非常值得研究的方向。在线社交网络服务是一种将云计算的资源分配和社交网络中用户关系数据资源分配相结合的产物,基于云的在线社交网络根据用户之间的关系进行合理的资源配置并提高QoS目标是主要的任务。在线社交网络在近些年取得了飞速的发展,在世界范围内拥使用社交软件进行信息交流的用户有数十亿,海量的数据需要存储,但是由于云环境中资源是按使用付费的,因此,如何合理的为每个用户分配资源对社交网络服务供应商来说是急需解决的问题,同时,如何将这些数据放在合适的位置减少用户的访问时间以提高用户使用体验也是十分重要的问题。针对这一问题,社交网络服务供应商可以通过智能算法为每个用户分配合理的资源,早期的云环境由于规模小,社交网络服务供应商最关注的是数据存储费用的问题,因此早期云环境中的数据放置算法优化目标为降低数据的存储费用。随着社交网络与云计算的不断发展,不仅用户对社交体验的要求越来越高,而且社交网络服务供应商也需要将放置算法的优化目标转移至降低多服务器之间的传输负载,提高社交网络中数据响应能力。近年来随着在线社交网络与云计算的不断发展,使得社交网络中的数据放置算法需要同时针对数据存储费用和多服务器之间的传输负载以及存储负载进行优化。因此,如何将数据存储费用和多服务器之间的传输负载以及存储负载三个目标有效结合,进而形成合适的优化目标又成为了当下研究的热点。目前在线社交网络常用的数据放置智能优化算法为遗传算法(Genetic Algorithm,GA),遗传算法是从一个初始种群开始进化搜索,因为不受具体的适应度函数影响,从多个点开始在全局以一定的概率并行搜索最优解,这样操作的有点是避免需要解决的问题陷入局部最优解。遗传算法会把需要解决的问题的参数表示成有限长度的染色体,并以一定的概率对染色体的形式进行选择、交叉和变异操作,而不是对问题本身进行求解,具有良好的可操作性。与传统的智能优化算法相比因为遗传算法可以在所有解的范围内进行搜索,因此遗传算法具有良好的鲁棒性、自适应性和全局寻优性。本文的主要工作如下:1.针对目前在线社交网络数据存储费用耗费巨大的缺点,利用遗传算法对传统的数据放置策略进行优化,在保证用户体验的前提下降低在线社交网络服务供应商存储数据所消耗的费用,减少成本。2.在优化数据存储费用的前提下降低多服务器之间的数据传输量,降低服务器之间的传输负载,通过实验证明该算法针对在线社交网络中的数据存储以及数据传输量的问题能够起到很好的优化效果,可以为社交网络服务供应商提供了更好的数据管理方法。3.在优化多服务器之间数据传输量的基础上优化服务器存储的负载均衡,消除服务器之间存储负载不平衡的情况,从而提高系统的稳定性。本文针对当前在线社交网络中的数据放置所存在的问题与不足进行了深入的探讨,利用智能算法对社交网络数据放置优化目标中较为重要的目标费用、数据传输量以及负载均衡分别进行了研究,最终通过实验证明该算法能够有效地降低社交网络中数据的存储费用和多服务器之间的数据传输量以及优化服务器的负载均衡。更多还原

【Abstract】 Cloud computing is a model of computing for resource sharing which based on Internet service.Cloud computing platform focus on storage and data processing,which provides resource of storage and computing to user on-demand while sharing massive distributed resources.The online social network platform is a kind of service platform that storing and processing by computer.In online social networks,service providers need to allocate resource with reasonable for each user according to the connection and location between users.With the development of social networking platform in cloud,the requirement of resource scheduling and quality of service constantly improving.Therefore,how to make the allocation scheme of resource more reasonable is an important research direction in cloud.The online social network service is a product that combines the resource of storage and computing in cloud.It is a major task for online social networks,which make the resource allocation reasonable and improve the quality of service.In recent years,the online social networks has achieved development rapidly.As the number of user increases,massive data needs to store.Thus,how to allocate resource of storage for every user is an urgent problem for social network service providers.As the same time,it is also important to store these data reasonable to reduce the user’s access time and improve user’s social experience.In order to solve the problem,social network service providers can allocate resources reasonable for each user with intelligent algorithms.Due to the small size of early cloud,the problem of social network service provides most concerned is the cost of data storage.Therefore,the target of data placement optimal algorithm is to reduce the cost of data storage in early cloud.With the development of social networks and cloud computing,not only the requirement of social experience get higher,but also the social network service provider turn the object of optimization to reducing the transmission load between multiple servers.Therefore,how to combine the three goals of data storage cost,transmission load and storage load effectively has become the focus of current research.Intelligent optimization algorithm for data placement used in online social networks is genetic algorithm commonly.Genetic algorithm is an evolutionary search starting from an initial population.Due to it has no affected by the specific fitness function,searching the optimal solution in parallel with a certain probability that start from a number of points.The advantage is to avoid getting into a locally optimal solution.Genetic algorithm could select,cross,and mutate the chromosomes at a certain probability.Due to the genetic algorithm can extend the search space to the whole problem space,and it has global optimization performance.Compared with the traditional optimal algorithm,the genetic algorithm has good robustness,adaptability and global optimization.The main work of this article is as follows:1.In the light of the disadvantages of the online social network data storage cost to optimize the traditional data placement strategy by using genetic algorithm.On the premise of ensuring the user experience,reducing the storage cost of social network providers.2.Reducing the transmission load between servers on the premise of optimizing the cost of data storage.Experiments demonstrate that the algorithm can optimize the data storage and data transmission in the online social network.It can provide better data management methods for social network service providers.3.Based on optimizing the data transmission volume between multiple servers,we optimize the load balance of server storage and eliminate the imbalance of storage load between servers.So as to improve the stability of the system.This paper makes an in-depth study of the data placement problem in the current online social network.The intelligent algorithm used to study the more important target of storage cost,data transmission and load balance in social network data placement.Finally,experiments demonstrate that the algorithm can reduce the data storage cost,the data transmission and load balance effectively between multiple servers in the social network.更多还原