摘要
1.引言目前的商用网管平台,如IBM的TivoliNetView/6000、HP的OpenView,其故障管理上的实现方式均是通过配置窗口将管理员的经验转变成一系列规则集,然后通过内部解析设定规则集中包含的若干MIB(管理信息库)变量的阈值,一旦监测进程或驻留在被管对象里的代理进程发现超过阈值,就以陷阱方式向网管站报警,内置的故障管理模块对陷阱信息经过过滤和关联后定位和鉴别故障原因,并启动预定制的恢复程序。尽管传统的故障管理系统能够在一定程度上发挥作用,但仍然存在以下弊端:
With the growing of network complexity and scale,fault modeling is becoming more difficult due to the dynamic nature and heterogeneity of network. We propose an intelligent agent for fault detecting based on adaptive learning algorithm. By segmentation measurement, MIB variable for describing network normal behavior is extracted and the deviation is detected. This information is combined in the structure of a Bayesian Graph so as to identify unknown or unpredictable faults.
出处
《计算机科学》
CSCD
北大核心
1999年第4期46-49,共4页
Computer Science
基金
国家"九五"重点科技攻关项目基金(96-743-01-01-02)
关键词
IP网络
故障监测
智能代理
网络管理
Network fault management, SNMP, MIB, Bayesian Graph, Adaptive learning algorithm. Intelligent agent
