复杂信息网络中的故障诊断与定位技术研究

发布时间：2018-02-14 06:20

本文关键词： 网络故障诊断探测技术故障定位网络流量综合探测　出处：《西安电子科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：网络故障诊断是网络管理的一个重要部分,对于网络服务质量的保证至关重要。对近几年常见的网络故障诊断方法进行了总结。将这些方法分为探测技术和数据处理方法两个大类并对它们进行了介绍说明,并对它们的适用范围和优缺点进行了分类比较。网络不断地发展对故障诊断技术提出了新的要求,对各类方法未来可做的方向和改进思路进行了讨论分析与展望。基于探测技术,提出了一种针对流量拥塞点的定位方法,能够更加清楚地了解节点的健康程度。对预计划与适应性两种探测方法的优势相结合,提出了一种新的综合式探测方案,使得探测所需的花费与探测时间均得到优化。此外还提出了一种探测间隔的动态调整策略,更加灵活合理地调整探测的时间间隔。现有方法多将节点分为健康与故障两种状态。然而,不同于故障节点,拥塞节点虽然并未完全停止工作,却严重影响了网络的传输性能。为了定位此类节点,提出一种基于流量值的网络故障诊断模型。模拟网络的流量传输模式,将节点状态划分的更为详细。探测站点的位置选取直接影响到所需站点的个数,与探测的花费值。提出了一种贪婪算法,迭代选取站点位置,使得在选取较少站点的情况下依然能达到较高的监测率。提出了监测路径的约减策略有效减少了监测花费。提出了监测路径选取策略,以及完整的拥塞节点、故障节点定位算法。实验仿真证明算法能够准确高效地定位网络中的拥塞节点及故障节点。探测技术中预计划探测与适应性探测有各自的优缺点,预计划探测诊断速度快,适应性探测花费低,结合两种探测的优点提出一种综合式探测方案。算法分故障探测及故障定位两个阶段,两阶段均按照预计划的思想发送固定的探测,而第二阶段是否执行则按照适应性探测的思想由第一阶段的结果决定。实验数据表明算法可在较低花费的情况下又以较快的速度定位故障。提出一种探测间隔时间的动态调整策略,根据网络不同时刻的健康状态动态调整诊断过程的间隔,在网络状态较好时增大探测间隔时间从而减少探测的花费,网络状态较差时缩短间隔保证更快速及时地定位故障节点,使得诊断行为更加灵活、合理。
[Abstract]:Network fault diagnosis is an important part of network management. This paper summarizes the common methods of network fault diagnosis in recent years. These methods are divided into two categories: detection technology and data processing method. The application scope, advantages and disadvantages of them are classified and compared. New requirements for fault diagnosis technology are put forward by the continuous development of network. In this paper, the future direction and improvement ideas of all kinds of methods are discussed and analyzed. Based on the detection technology, a new location method for traffic congestion points is proposed. It is possible to understand the health of nodes more clearly. A new integrated detection scheme is proposed to combine the advantages of pre-planning and adaptive detection methods. The cost and time of detection are optimized. In addition, a dynamic adjustment strategy of detection interval is proposed. The existing methods divide the nodes into health and fault states. However, unlike the fault nodes, the congested nodes do not stop working completely. In order to locate such nodes, a network fault diagnosis model based on traffic value is proposed. The location selection of the probe site directly affects the number of the required sites and the cost of the probe. A greedy algorithm is proposed to iteratively select the site location. Under the condition of selecting fewer stations, the monitoring rate can still reach a higher rate. A reduction strategy of monitoring path is proposed, which can effectively reduce the monitoring cost, and a monitoring path selection strategy is proposed, as well as a complete congestion node. The simulation results show that the algorithm can accurately and efficiently locate the congestion nodes and fault nodes in the network. The pre-plan detection and adaptive detection have their own advantages and disadvantages, and the pre-plan detection and diagnosis speed are fast. The adaptive detection cost is low, combining the advantages of the two kinds of detection, a comprehensive detection scheme is proposed. The algorithm is divided into two stages: fault detection and fault location, both of which send fixed detection according to the idea of pre-planning. The second stage is determined by the result of the first stage according to the idea of adaptive detection. The experimental data show that the algorithm can locate the fault at lower cost and at a faster speed. A detection interval is proposed. Dynamic adjustment strategy based on. According to the health state of the network at different times, the interval of diagnosis process is dynamically adjusted, the detection interval time is increased when the network is in good condition, and the cost of detection is reduced. When the network condition is poor, shortening the interval ensures that the fault node can be located more quickly and timely. The diagnosis behavior is more flexible and reasonable.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP393.06

【相似文献】