分布式RDF数据管理综述

发布时间：2019-04-28 20:36

【摘要】：资源描述框架(resource description framework,RDF)作为一个展示、共享和连接网络上的数据的模型,已经被广泛地用在各种应用中.同时,SPARQL(simple protocol and RDF query language)作为一种结构化查询语言则被用来支持对RDF数据进行查询检索.随着RDF数据规模的日益增长,在现有RDF数据库上进行SPARQL查询处理已经超出了单机的处理能力.于是,人们需要设计出高性能的分布式RDF数据库以支持对SPARQL查询进行高效的处理.当前,已经有大量的工作来讨论如何搭建分布式RDF数据管理系统.对这些不同的分布式RDF数据管理方法进行综述,将现有的分布式RDF数据管理方法分成3类:基于云计算平台的分布式RDF数据管理方法、基于数据划分的分布式RDF数据管理方法和联邦式系统.基于云计算平台的分布式RDF数据管理方法利用已有云平台进行RDF数据的管理;基于数据划分的分布式RDF数据管理方法首先将RDF数据图划分成若干子图,然后将这些子图分配到不同计算节点上;联邦式系统的特点是数据已经分布在不同节点上,数据管理系统无法控制数据的分布.在每类分布式RDF数据管理方法的介绍中,将深入讨论以帮助读者了解各种方法的特点.
[Abstract]:Resource description Framework (resource description framework,RDF), as a model for displaying, sharing and connecting data on the network, has been widely used in a variety of applications. At the same time, as a structured query language, SPARQL (simple protocol and RDF query language) is used to support query retrieval of RDF data. With the increasing scale of RDF data, SPARQL query processing on the existing RDF database has exceeded the ability of single machine processing. Therefore, people need to design a high-performance distributed RDF database to support efficient processing of SPARQL queries. At present, a lot of work has been done to discuss how to build a distributed RDF data management system. This paper summarizes these different distributed RDF data management methods, and divides the existing distributed RDF data management methods into three categories: distributed RDF data management methods based on cloud computing platform, and distributed RDF data management methods based on cloud computing platform. Distributed RDF data management method and federated system based on data partition. The distributed RDF data management method based on cloud computing platform makes use of the existing cloud platform to manage the RDF data. The distributed RDF data management method based on data partition firstly divides the RDF data graph into several sub-graphs and then allocates these sub-graphs to different computing nodes. The characteristic of federated system is that the data is distributed on different nodes, and the data management system can not control the distribution of data. In the introduction of each kind of distributed RDF data management methods, it will be discussed in depth to help readers understand the characteristics of various methods.
【作者单位】：北京大学计算机科学技术研究所;湖南大学信息科学与工程学院;
【基金】：国家自然科学基金优秀青年科学基金项目(61622201)~~
【分类号】：TP311.13

【相似文献】