动态反馈的异构集群负载均衡算法的实现
CSTR:
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

基金项目:

教育部留学回国人员启动基金资助项目[2007110810]


Research and realization of the load balancing algorithm for heterogeneous cluster with dynamic feedback
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    虚拟服务技术(LVS)的集群负载调度系统中的加权调度算法权值是静态的,没有动态调整机制,不能依据真实服务器处理能力进行动态的任务分配;给出了一种通过量化调度器和真实服务器的实时反馈负载信息指标,采用服务器负载率和分配权值为计算指标,引入临界回归因子,利用动态反馈临界加速回归的算法思想,实现基于动态反馈机制的动态反馈临界加速回归分配算法。该算法能根据负载均衡调度器和业务处理服务器的实时反馈信息,及时进行负载调度,实现负载动态平衡,进一步提高服务器的利用效率和集群系统的吞吐率,并在LVS 的负载平衡调度集群系统中进行了应用性的实验验证。测试结果表明,该算法可实时反馈负载信息,动态进行负载调度,整体上较好的实现了负载动态平衡,提高了服务器的利用效率和集群系统的吞吐率。

    Abstract:

    Linux virtual server (LVS) can establish a server cluster based on Linux with high availability, high capability, high reliabilities and scalability, while the weights of the weighted scheduling algorithm for the cluster load balancing system in LVS is static. Without the dynamic adjusting mechanism, the task can not been assigned dynamically according to the servers’ real capabilities. A dynamic load balancing algorithm (Multiplicative Regression in Critical Area,MRC) is introduced to adjust weights of the servers dynamically. The MRC algorithm calculates the load ratios and distributed weights according to the realtime feedback information from the real servers, and introduces the threshold regression factors to schedule tasks for web server cluster to achieve load balance. The method improves server's efficiency and the capability of the whole system. The method is applied to web server cluster systems based on Linux virtual server. The experimental results show that the algorithm can achieve realtime feedback of the load information, dynamic load scheduling, and achieve dynamic load balance in a whole, which improves the server utilization efficiency and the capability of the whole system.

    参考文献
    相似文献
    引证文献
引用本文

陈伟,张玉芳,熊忠阳.动态反馈的异构集群负载均衡算法的实现[J].重庆大学学报,2010,33(2):73-78.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2009-10-12
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期:
  • 出版日期:
文章二维码