一种基于动态贝叶斯模型的大型IP网络故障诊断方法

为了提高IP业务的服务质量,利用告警等症状和已有知识快速准确地定位根故障十分重要。基于贝叶斯网络的不确定推理方法是近年来广泛应用的一种故障诊断方法。目前,基于静态贝叶斯网络的故障定位只是利用当前信息进行故障诊断,无法处理时间信息;而已有基于动态贝叶斯网络的诊断算法复杂度太高,不适用于大型网络。本文针对大型IP网络,建立用于故障诊断的动态贝叶斯模型,并对基于动态贝叶斯网络的一种通用的精确算法进行改进,实验证明它能够对大型IP网络快速准确的定位故障。本文方法充分利用告警库中的历史数据和当前症状信息,对当前的系统状态进行估计,完成故障诊断。

Fault Diagnosis for Large-scale IP Networks Based on Dynamic Bayesian Model

To improve the quality of IP service, it is important to quickly and accurately diagnosis the root fault from the observed symptoms and knowledge. The approximate inference based on Bayesian networks is the most popular fault diagnosis technology in recent years. Presently, fault localization based on Bayesian networks is only according to the current information and does not consider the time information. The existing methods based on dynamic Bayesian networks aren＇t fit for large-scale networks because of their complexity. This paper establishes a fault diagnosis model for large-scale IP networks based on dynamic Bayesian networks, at the same time improves a universal exact algorithm and implements simulation. The results show that the algorithm can run well. This method makes full use of the historical data and current observations to estimate the current system state and complete the fault diagnosis.