基于多Agent的分布式IP网故障诊断系统设计与实现

摘要

为便于及时定位并排除典型网络故障，以数据包分析为基础，基于多Agent系统，设计实现了一种便于扩展的分布式IP网故障诊断系统。该系统采用网络端口镜像的方式进行数据采集，采用伯克利数据包过滤器(BPF)包过滤机制进行数据包过滤，基于专家系统进行典型网络故障诊断。此外，系统采用三元组＜实体，属性，规则集＞进行知识表示，用支持度表示规则出现的频率。实际应用表明，该系统可有效地帮助网络维护人员及时了解网络状态，快速、准确地诊断典型网络故障。%To locate and eliminate typical network faults in time,the paper designs an expandable distributed IP network fault diagnosis system based on data packet analysis and multi-agent system. The system adopts the method of network port mirroring to collect data,conducts data filtering with Berkeley Packet Filtering (BPF)mechanism,and performs network fault diagnosis based on expert system.In addition,the system uses tripleto realize knowledge repre-sentation,and represent the frequency of the rule appearance by support degree.The practical appli-cation shows that the system can effectively help network maintenance personnel to know the network status in time and quickly and accurately diagnose typical network faults.