The Research Of Semi-Online Active Probing Algorithm For Fault Diagnosis

Fault management is one of the basic functions in network management, which aims to ensure the reliability of the network system. When network management system detects the occurrence of faults in the network, network fault management system has to locate the dysfunction part, lower the damage made by the failure, assure the availability and reliability of the network. Nowadays, the network fault management system usually listens to the network passively, analyzes the collected alerts with event correlation technology to locate the dysfunction part. Network management system of this kind need the network element with interfaces to report alerts. Also the massive redundant alerts are a great burden for the network management system. Network management systems of this kind are not suitable for the network due to the growing complex network. Fault localization based on active probing, which sends detective sequences actively to get the state of the network, is a highly efficient adaptive method which could detect and locate faults in real-time. Active probing is prospective in fault management due to it is activeness and adaptive which could offset the disadvantages of alarms.This thesis studies the basic methods of fault localization using active probing, analyzes and improves the existing methods, proposes more efficient algorithms, evaluates the applicability of the methods with emulation, illustrates how to apply active probing in fault management with a real network fault management system. The contents of this paper include:1) Study and summarize the backgrounds and current status of fault management 2) Study and summarize the general methods and key problems of using active probing in fault management, including process and algorithm.3) Analyze and improve probe selection algorithm in fault localization part, and propose a highly efficient active probing algorithm 4) Use this algorithm to solve the problem of fault detection under different models.5) Verify the effectiveness of the algorithms through simulation experiments and analyze the experiments'results.6) Illustrates the applicability of active probing in a fault management system.