DOA下的数据注册中心元数据存储策略研究

发布时间：2018-05-20 02:23

本文选题：DOA + 元数据存储　；参考：《成都理工大学》2013年硕士论文

【摘要】：伴随国家信息化建设的快速发展,各种各类数据信息不断涌现,时时刻刻为人们的日常生活生产提供信息服务。Internet的全世界普及,使数据每一时刻都在网络中流动,使数据散落在不同的数据存储系统与设备中。为实现数据在客户端汇聚,G/S模式应运而生。它采用“请求(request)-聚合(Polymerization)-服务(Service)"机制对所有的数据信息服务过程进行组织与管理,在客户端实现数据信息汇聚和服务聚合。G/S模式解决了数据的汇聚问题,但是如何管理各种数据,G/S模式并没有提出有效的解决方法。为了能够更快更好的获取数据,为G/S模式服务,成都理工大学教授苗放老师提出了以数据为中心,以标识为主线的面向数据的体系结构——DOA,运用数据的元数据实现对数据的统一管理。要管理各种类型数据,首先要有相关数据的元数据,所以数据的元数据存储问题,成为了DOA首要考虑的问题。DOA要求数据注册中心能够快速提供数据的元数据信息,如何存储才能快速提供元数据,成为DOA下元数据管理数据的瓶颈,也是本文研究的问题,解决DOA下的数据注册中心元数据快速获取的存储问题,成为当务之急。本文首先对G/S模式和DOA进行了全面概述,分析了它们的层次结构与对数据汇聚和管理的作用,促使了对元数据存储策略的研究。通过对元数据标准和开发设计框架基础知识的概述,为设计了元数据存储规范准备了基础知识。本文的工作如下：(1)研究数据注册中心元数据存储策略由于DOA使用元数据统一管理各种数据,所以元数据的获取速度决定了DOA的性能。因而,如何存储元数据成为元数据快速获取的关键。在分析研究文件系统方式、LDAP目录存储方式、各种关系数据库存储方式和内存数据存储方式的存储优缺点的基础上,对使用何种存储方式才能使数据注册中心元数据被快速获取进行了分析研究。并依据研究成果,设计了一种能够实现元数据快速获取的元数据存储结构模型。(2)研究元数据存储的相关算法基于内存容量的大小和元数据被访问频率的问题,研究为内存提供存储元数据的相关算法。研究了一种基于代价函数的缓存替换算法,为数据注册中心在缓存中动态存储最近有可能被经常访问的元数据。同时进一步研究了在缓存中选取高访问频率元数据表的算法,以实现通常访问的元数据信息存入内存数据库。通过这两种算法,把访问频率高的元数据存入内存,提高了数据注册中心对元数据的获取速度。本文的创新点与研究成果有：(1)提出了一种DOA下的数据注册中心元数据快速获取存储策略。本元数据存储策略,依据DOA下的元数据统一管理数据的要求,利用关系数据库的存储容量、内存数据库的访问效率和缓存动态存储作用,通过基于代价的缓存替换算法实现在缓存中动态存储最近有可能被经常访问的元数据,以提高元数据的获取速度与积累元数据的被访问频率；再通过基于代价函数的缓存选取算法,选取缓存中访问频率最高的元数据表存入内存数据库,完成元数据的内存数据库存储。在此基础上,设计了DOA下的数据注册中心的元数据存储结构模型,并依据此模型设计了DOA下的数据注册中心元数据存储系统。(2)提出了一种数据注册中心的替换缓存中访问权值最低元数据表的算法。通过对现存的几种常用缓存替换算法分析研究,从函数的角度,利用缓存数据表大小、访问次数、访问时间和获取缓存数据表的CPU时间,设计了一种基于代价函数的缓存替换算法,实现最近有可能被经常访问的数据表在缓存中的存储。(3)提出了一种数据注册中心的选取高访问频率元数据表存入内存数据库的算法。在对现存缓存替换算法研究的基础上,从函数的角度,利用缓存数据表大小、访问次数、访问时间和访问密度,设计了一种基于代价函数的缓存数据表选取算法,实现高访问频率元数据表从缓存到内存数据库的存储。
[Abstract]:With the rapid development of national information construction, various kinds of data and information are constantly emerging. The information service of.Internet is popularized all the time for people's daily life. The data will flow in the network at every moment, so that data are scattered in different data storage and storage systems and equipment. G/S mode comes into being. It uses "request" - aggregation (Polymerization) - Service (Service) "mechanism to organize and manage all the data information service processes. The data aggregation problem is solved by data aggregation and service aggregation.G/S mode in the client, but how to manage all kinds of data, G/S mode is not In order to get the data faster and better, to serve the G/S mode faster and better, Professor Miao Fang of Chengdu University of Technology put forward a data centric, identification oriented data oriented architecture, DOA, using data metadata to realize the unified management of the logarithm data. First, to manage various types of data, first In order to have metadata of relevant data, the problem of metadata storage of data has become the primary concern of DOA..DOA requires that data registration center can provide metadata information quickly, how to store metadata quickly, become the bottleneck of metadata management data under DOA, and the problem of this paper, to solve DOA The fast acquisition of data registry metadata has become a top priority. This paper first gives a comprehensive overview of the G/S model and DOA, analyzes their hierarchical structure and the function of data convergence and management, and promotes the research of metadata storage strategy. The basic knowledge is prepared for the design of metadata storage specification. The work of this paper is as follows: (1) study the metadata storage strategy of data registration center because DOA uses metadata to manage all kinds of data, so the speed of metadata acquisition determines the performance of DOA. Therefore, how to store metadata is the key to the rapid acquisition of metadata. On the basis of analyzing the methods of file system, the storage mode of LDAP directory, the storage mode of all kinds of relational databases and the memory of memory data storage methods, this paper analyzes and studies what kind of storage methods used to make the metadata of the data registration center be obtained quickly. And a kind of realization is designed based on the research results. The metadata storage structure model is quickly acquired. (2) the related algorithms of metadata storage are based on the size of memory capacity and the frequency of the metadata being accessed, and the related algorithms for storing metadata for memory are studied. A caching replacement algorithm based on cost function is studied, which is used to move the data registry in the cache. State storage of metadata that may be frequently accessed recently. At the same time, the algorithm of selecting high access frequency metadata in the cache is further studied to save the commonly accessed metadata information into the memory database. Through these two algorithms, the metadata with high frequency of access is stored in the memory, and the data registry is improved to the metadata. The innovation points and research results of this paper are as follows: (1) a fast acquisition strategy for data registration center metadata under DOA is proposed. This metadata storage strategy is based on the requirements of unified management of data under the metadata under DOA, using the storage capacity of the relational database, the access efficiency of the memory database and the dynamic storage of the cache. Use the cost based cache replacement algorithm to dynamically store the most frequently accessed metadata that may be frequently accessed in the cache to improve the speed of metadata acquisition and the access frequency of the accumulated metadata; then, the cache selection algorithm based on the cost function is used to select the metadata with the highest access frequency in the cache and save the memory. On the basis of this, the metadata storage structure model of the data registration center under DOA is designed, and the data registration center metadata storage system under DOA is designed based on this model. (2) a calculation of the minimum metadata table of access weight in the replacement caching of the data registration center is proposed. By analyzing several existing common cache replacement algorithms, from the point of view of function, using the size of the cached data table, the number of access, access time and the CPU time of obtaining the cached data table, a cache replacement algorithm based on the cost function is designed to realize the storage of the data table that may be frequently accessed recently in the cache. (3) an algorithm for selecting the high access frequency metadata table to be stored in the memory database is proposed. On the basis of the existing cache replacement algorithm, a cost function based caching data table is designed from the function point of view, using the size of the cached data table, access times, access time and access density. The algorithm implements the storage of high access frequency metadata tables from cache to memory database.
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP333

【相似文献】