北京邮电大学学报

  • EI核心期刊

北京邮电大学学报

• 论文 • 上一篇    下一篇

电信社群网络中介度的网格并行算法及调度算法

陈 平, 王 柏, 徐六通, 吴 斌, 王艳辉   

  1. 北京邮电大学 通信软件工程中心 北京 100876
  • 收稿日期:2006-01-25 修回日期:1900-01-01 出版日期:2006-05-30 发布日期:2006-05-30
  • 通讯作者: 陈 平
  • 基金资助:
     

Grid Based Parallel and Schedule Algorithm for Betweenness Computation in Telecom Social Network Graph

CHEN Ping, WANG Bai, XU Liu-Tong, WU Bin, WANG Yan-Hui   

  1. Telecom Software Engineering Center, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2006-01-25 Revised:1900-01-01 Online:2006-05-30 Published:2006-05-30
  • Contact: CHEN Ping
  • Supported by:
     

摘要: 为了解决电信社群网络中介度(图的一个几何量)计算中的海量计算问题,研究并实现了高性能网格并行计算方法。该算法采用层次性的二分法分割数据,能在较短的时间内完成大规模社群网络图的各个顶点中介度计算。为了提高计算的加速比,还提出了一种改进的网格并行调度算法,采用动静态结合的方法来平衡负载。论证表明,改进算法的加速比、并行效率和平衡度都有提高,计算用时与网格上并行计算处理器数目成近似线形关系。

关键词: 复杂网络, 电信社群网络, 中介度, 网格并行, 任务调度算法

Abstract: For betweeness computation in Telecom Social Network graph with huge amount of data, a grid based parallel algorithm is presented. It adopts the hierarchic dichotomy of data. Meanwhile a new schedule strategy is proposed, that combines static and dynamic methods to reach load balance. The experimental results justified that the algorithm achieves a higher speedup ratio and the parallel efficiency, and the time consumption is approximately proportion to the number of parallel CPUs as well.

Key words: complex network, telecom social network, betweeness, grid parallel computation, task scheduling algorithm

中图分类号: