服务库多级索引模型性能分析

发布时间：2018-01-17 14:41

本文关键词：服务库多级索引模型性能分析　出处：《江苏大学》2017年硕士论文　论文类型：学位论文

【摘要】：面向服务的计算(Service-Oriented Computing,SOC)为企业软件集成和分布式系统提供了一种灵活的解决方案。面向服务的体系架构(Service-Oriented Architecture,SOA)是面向服务计算的基础设施,其基本的要素是Web服务。在最近几年,Web服务的数量急剧增加,同时快速发展的云计算也促进了服务的发展,大量的服务汇集于以云平台为中心的服务库。服务发现和服务组合是当前服务计算领域的两个重要研究方向。因此,如何高效地管理一个大规模的服务库并能有效促进服务发现和服务组合将是一个非常重要且具有挑战性的问题。本文首先从服务的研究背景及意义入手,研究与介绍了服务发现和服务组合的研究现状,然后在此基础上进一步研究了服务索引结构,对倒排索引以及多级索引模型的三种部署模型进行了较为详细的研究,具体的研究内容包括:(1)目前大规模服务库主要是以倒排索引的结构进行存储,虽然倒排索引可以在很大程度上缩小服务发现和组合的查找空间,但仍然存在大量的信息冗余,导致效率低下。针对倒排索引信息冗余问题,多级索引模型以及其三种部署模型的存储结构在最大程度上消除了信息冗余问题,并能有效促进服务发现和服务组合效率。但针对特定的服务库,多级索引模型并没有给出一个用来选取其效率最高的索引模型的有效标准,针对这个问题,本文首先对不同索引模型结构进行了性能分析,并根据服务库以及服务的特点给出了期望函数,作为服务库选取高效服务索引模型的依据。然后给出了详细的五种索引模型的期望函数的公式推导过程,并通过优化部署方案分析了各个索引模型的特点以及其适用的服务库的特点等。最后,实验结果验证了所提优化部署方案以及期望函数的准确性和可靠性。(2)针对多级索引模型中的键值优化选择算法进行了初步探讨。首先,对多级索引模型中的键值选取原理进行了深入的分析与介绍,并通过多种简单示例帮助理解。然后分析了键值选取中存在的问题。最后,通过实验对比键值优化与非优化状态下索引结构的检索性能,说明了键值在随机选取情况下并没有对索引结构的检索性能造成任何影响,证明了键值优化选取中存在的问题,也证实了对键值优化选择问题的初步探讨的正确性。
[Abstract]:Service-Oriented Computing. Service-Oriented Architecture provides a flexible solution for enterprise software integration and distributed systems. SOA) is the infrastructure of service-oriented computing, whose basic element is Web services. In recent years, the number of web services has increased dramatically, and the rapid development of cloud computing has also promoted the development of services. A large number of services are gathered in the cloud platform-centric service library. Service discovery and service composition are two important research directions in the field of service computing. How to efficiently manage a large-scale service library and effectively promote service discovery and service composition will be a very important and challenging issue. This paper starts with the research background and significance of service. This paper studies and introduces the research status of service discovery and service composition, and then further studies the service index structure. In this paper, three deployment models of inverted index and multilevel index model are studied in detail. The specific research contents include: 1) at present, the large scale service library is mainly stored in the structure of inverted index. Although inverted index can reduce the search space of service discovery and composition to a large extent, there is still a lot of information redundancy, which leads to inefficiency. Multi-level index model and its storage structure of three deployment models can eliminate the problem of information redundancy to the maximum extent, and can effectively promote the efficiency of service discovery and service composition. However, the storage structure of the multi-level index model and its three deployment models can effectively promote the efficiency of service discovery and service composition. The multilevel index model does not give an effective standard to select the most efficient index model. In order to solve this problem, this paper first analyzes the performance of different index models. According to the characteristics of the service library and the characteristics of the service, the expected function is given as the basis for the service library to select the efficient service index model. Then, the derivation process of the expectation function of the five index models is given in detail. And through the optimization of deployment scheme analysis of the characteristics of each index model and its applicable service library characteristics and so on. Finally. The experimental results verify the proposed optimal deployment scheme and the accuracy and reliability of the expected function. The selection principle of key and value in multilevel index model is deeply analyzed and introduced, and some simple examples are used to help us understand. Then, the problems in key value selection are analyzed. Finally. By comparing the retrieval performance of the index structure in the state of optimization and non-optimization, it is shown that the key value has no effect on the retrieval performance of the index structure in the case of random selection. The problems in the selection of key and value are proved, and the correctness of the preliminary discussion on the selection of key and value is also confirmed.
【学位授予单位】：江苏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【相似文献】