面向数据高交互任务的分布式图计算方案的设计与实现  

Design and Implementation of Distributed Graph Computing Scheme for High Data Interaction Tasks

作　　者：俞山青[1] 王甬琪 崔文豪 孟栎均 李冰 傅晨波[1] YU Shan-qing;WANG Yong-qi;CUI Wen-hao;MENG Li-jun;LI Bing;FU Chen-bo(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310000,China;Hangzhou Zhongao Technology Company Limited,Hangzhou 310000,China)

机构地区：[1]浙江工业大学信息工程学院,杭州310000 [2]杭州中奥科技有限公司,杭州310000

出　　处：《小型微型计算机系统》2020年第10期2098-2103,共6页Journal of Chinese Computer Systems

基　　金：浙江省基础公益研究计划项目(LGF20F02001)资助。

摘　　要：对于较大规模的复杂图计算任务,集中式计算方案往往难以胜任,而现有的分布式图计算工具又会由于节点间数据传输耗时过多而效率低下.基于这种情况,本文提出了一种面向数据高交互任务的分布式图计算方案,用于提高此类任务的计算性能.该方案基于任务分割的思想,将系统划分为任务管理中心、数据中心与计算节点.其中任务管理中心负责任务的分割、分发、定时状态检测、汇总等管理工作;计算节点完成分割后子任务的计算;数据中心对图数据及计算结果进行保存与管理.此外,本文以图中心性计算为目标,实现了上述分布式图计算方案并进行相关实验.实验结果表明,本文所设计的方案可以有效提升此类数据高交互图计算任务的计算效率.For large-scale complex graph computing tasks,centralized computing schemes are often hard to accomplish efficiently,and the existing distributed graph computing schemes are also inefficient due to excessive data transmission betw een calculating nodes.Based on this situation,this paper proposes a distributed graph computing scheme for high data interaction tasks,w hich is used to improve the computational performance on such tasks.Based on the idea of task segmentation,the scheme divides the system into tasks management center,data center and computing nodes.The tasks management center realizes the management of tasks,w hich includes task segmentation,tasks distribution,tasks state detection and tasks aggregation.The computing nodes realize the calculation of the subtasks.The data center is used to save and manage the graph data and the calculation results.In addition,the paper realized the abovementioned distributed graph calculation scheme based on centrality calculation on vertex and carried out related experiments.The experimental results show ed that the scheme designed in this paper could improve the computational efficiency on such high data interaction graph computing tasks effectively.

关 键 词：数据交互 分布式 图计算 中心性计算 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]