华东师范大学学报(自然科学版)

• 计算机科学 • 上一篇    下一篇

一种基于关系数据库的图计算平台

蒋 奎[1] , 陈 亮[2]   

  1. 1. 河北轨道运输职业技术学院, 石家庄 050021; 2. 微软亚洲研究院, 北京 100080
  • 收稿日期:2016-06-24 出版日期:2016-09-25 发布日期:2016-11-29
  • 通讯作者: 陈 亮, 男, 研究员, 研究方向为数据库及信息系统. E-mail: jeche@microsoft.com.

A RDBMS-based graph computing platform

JIANG Kui[1], CHEN Liang[2]   

  1. 1. Hebei Vocational College of Rail Transportation, Shijiazhuang, 050021 China; 2. Microsoft Research Asia, Beijing 100080, China
  • Received:2016-06-24 Online:2016-09-25 Published:2016-11-29

RichHTML

10

PDF

660

摘要：

本文提出了一种新的基于关系数据库管理系统 (Relational Database Management System, RDBMS)(本文简称关系数据库) 的图计算平台. 该平台将图数据以原生的形式在关系数据库的表格中存储, 从而在数据表达上和原生图计算平台达到了一致. 该平台将图计算逻辑完整准确地表达为 SQL (Structured Query Language) 查询语句. 关系数据库执行 SQL 查询语句, 从而完成图计算, 并将结果返回. 实验结果表明, 该新的平台有效地利用了关系数据库成熟的查询优化技术, 在很多方面优于现有的原生数据平台; 而目前的性能局限, 也会随着未来关系数据库的不断演化和迭代, 得到有效的解决.

关键词: 关系数据库管理系统, 图计算, 原生图计算平台

Abstract:

This paper proposes a new RDBMS-based (relational database management system) graph computing platform. In this platform, graph data is represented in native data structures, achieving the same representation as in native graph computing systems. On top of this native representation, graph algorithms are expressed as SQL (Structured Query Language) statements, which are executed by the underlying relational database systems. Experimental results show that this new graph computing platform leverages mature SQL technologies on query optimization and execution, thereby providing superior performance in many aspects. Its current performance limitations, on the other hand, will be overcome by future evolution and optimization of relational database systems.

Key words: RDBMS (relational database management system), graph computing, native graph computing platforms