【作者】 范清风；

【导师】 何炎祥； Ka. Zeitouni （Fench）；

【作者基本信息】 武汉大学 ， 计算机软件与理论， 2013， 博士

【摘要】 计算机的应用领域不断扩大,已经逐渐遍及人类生活的各个领域。目前,在越来越多的科学、生产、工程等领域,产生的大量数据,如何有效地管理和利用这些数据,已成为一个重要研究课题。网格(Grid)应运而生,数据网格(Data Grid)是网格技术的延展。本文主要聚焦于数据网格的海量数据传输策略研究。本文首先讨论了数据网格信息服务的改进策略。然后,讨论了在文件级以上环型复制策略和线型复制策略及文件级以下的文件分块复制策略,用以提高传输效率。接着,进一步讨论如何对数据网格系统地建立一个优化的组播体系结构,进而在根本上提高数据传输效率。最后,体现从定性到定量的思想,综合考虑节点位置因素W’和数据因素W”,构建线性方程组,解线性参数,从而构建优化的组播树。在以前工作的基础上,还提出了从二维到三维向量博弈均衡的扩展,以及扩展到弹性有限维。本文主要有以下贡献：1.用环型复制策略和线型复制策略提高数据网格信息、服务的效率。在数据网格环境中,数据网格信息服务是网格计算的基本服务,它是其它数据网格应用的基础,是数据网格研究的重点和难点。为了提高信息服务的效率,本文提出了环型复制策略和线型复制策略两种改进策略。根据不同的网格配置,与其它策略相比较,本文最终证明了环型复制策略和线型复制策略能够提高信息系统的服务效率。2.在文件级以上用环型,线型复制策略,在文件级以下用文件分块复制法提高数据网格的传输效率。为了提高系统效率,本文提出了在文件级以上应用环型策略和线型策略。在文件级以下用文件分块复制法把数据网格文件划分成很多块,在多个服务器结点之间并行复制。考虑到不同的网格基础,与其它的复制策略相比较,基于试验证明了我们的方法能够提高数据网格的传输效率。3.用博弈均衡级层组播树算法提高数据网格的效率在构建数据网格体系结构的时候,应该考虑节点位置因素,和节点数据量因素。在上一节的基础上进一步体现从定性到定量的思想,精确提高数据网格传输效率。而且,位置和数据两个因素彼此独立,而又相互关联,彼此博弈而又均衡。本文找到它们的博弈均衡点,进而解出线性参数。扩展的模拟结果证明,与其它现存的著名算法相比较,该算法更适合于数据网格。4.本文的组播算法可以从2维扩展到3维向量博弈均衡,及弹性有限维。我们的研究工作是一个有机的体系结构。是本着从低级到高级,从简单到复杂的原则,逐层深入,逐层扩展。第N步是源与第N-1步,还可以推出第N+1步。本章把我们的模型从二维扩展到第三维,进一步扩展到弹性有限维。因为在理论上影响网格无线组播通讯的因数是无限的,但是在实际应用中是有限的,3,4,5维就已经很复杂了。我们的算法根据已有的平台,做适当的扩展,可以无缝的整合多个因数,所以具有更强的普世性,和扩展性。这一点是其他算法所不具备的。关于将来的工作。可以对数据向量研究专门的子算法。本文中,位置向量是复杂而变化的,但数据权向量是简单而确定的。然而,在实际应用中,数据量会随着时间和条件而改变。所以可以对数据权向量研究专门的子算法。第三,讨论两个向量的非线性关系：目前本文仅仅讨论两个向量的线性关系,但是现实中,系统传输不是根据两个向量的线性关系,而是根据两个向量之间的非线性关系。最后,在讨论了两个向量的关系以后,本文很容易地扩展到3个向量的关系,甚至可以扩展到多个向量的关系。更多还原

【Abstract】 With the amount of data in many science fields dramatic increasing. At present, more and more science, production, engineering and other fields[1,2,5], etc, also the need of analyzing and dealing with the amount of data is also fast expanding. How to effectively manage and use the data has become an important research topic. The Grid technology came into being. Data Grid is an extension of the grid technology on data management which is more focused on data storage, transmission and management[1]. This paper mostly focuses on the research on the mass data transmission Strategy of Data Grid. The paper firstly introduces the improving strategy of data grid information service, then discusses over file degree the ring and thread replication strategy and over file degree the file-parted replication strategy to improve the efficiency of Data Grid. After that it discusses how to generate a multicast architecture for Data Grid. The system comprehensive consider about the place factor and data quantity factors, building multicast architecture, so radically improve the efficiency of data transmission of Data Grid. At last, according to the idea of from qualitative to quantitative, comprehensively think about the place factor and data quantity factor, introduce the idea of game theory, finding the game and balance point of the place factor and data quantity factor, moreover resolve the value of line parameter. Furthermore, after discussed two vectors correlation, the paper can easily extent to3weight vectors correlation, At last, the paper can extend to N vectors correlation.The contribution of this paper:1. Ring replication and thread replication improving the efficiency of information service.In Data Grid environment, Information Service is primary service, which is the base of other applications. Compared with many others replication strategies and considered in different grid infrastructures, the strategies of Ring replication and Thread replication can boost the efficiency of the Data Grid information service system.2. Over file degree using the ring and thread replication strategy and below file degree using the file-parted replication strategy to improve the efficiency of Data Grid.In Data Grid, it is inevitable that there is a mass of data. To improve the system efficiency, over file degree the paper puts forward the strategies of ring replication and thread replication and below file degree the paper will study the viewpoint of the file-parted replication to divide the data Grid files into several blocks that are replicated parallel between the Data Grid sever points. Finally, compared with many other replication strategies, our approachs can improve the efficiency of Data Grid.3. Using linear optimal hierarchical multicast tree algorithms to improve the efficiency of Data GridWhen Data Grid system builds architecture, it should comprehensively consider about the place factor and data quantity factor. On the base of last section, according to the idea from qualitative to quantitative, it accurately improves the efficiency of transmission of Data Grid. We find the game and balance point, then resolves the value of line parameters, construct multicast architecture, transfer data. Comparing with former approaches, our algorithm, extended simulation results indicate that the new algorithms are more suitable for Data Grid compared with other well-known existing multicast solutions.4. To extend2vectors correlation to3vectors correlation and extended to elastic finite-dimensional.Our research work is an organic architecture. N step is derived from the N-1step, but also launched the n+1step. In theory, the factors that affect the wireless multicast communication is infinite, which is discussed in the future work of our new paper. But in practice, we often utilize finite multi-dimensional,3,4,5, such as space, data, economy, politics, military, and so on. Based on the powerful computing capability of modern computer, the finite element program can resolve the algebraic equations with hundreds of thousands of degrees of freedom. About the future work:After that the spatial weight vector is complex, but the data quantity weight vector is definite. However, in some situation the data quantity could be changing, so the paper should study special sub-algorithm for it also.Moreover, in here the paper just talks about the liner relationship of two vectors. But, sometime, the system transforms according to the non liner relationship of two vectors.更多还原