Abstract: With the rapid development of telecom IPTV business, the scale of carrying network has constantly expanded, and the difficulty in operation and maintenance of network equipment has gradually increased. When some equipments become faulty, how to quickly locate the faulty nodes in a large-scale network has become a major challenge in operation and maintenance. At present, the equipment fault alarm of carrying network depends mainly on the equipment performance logs, which generates many false alarms, and can not adapt to the fault location in a large-scale network. Therefore, by using the tools such as Spark, the abnormal nodes location method based on terminal data is proposed, combining network topology to realize threshold alarm preliminarily. And further based on the analysis of equipment correlation and reliability, a large number of false alarms generated during the process of threshold alarm are cleaned to improve the precision rate of fault location. The experimental results show that this method has high practical value and its precision rate can reach 89% in the fault location of carrying network.

Key words: Internet Protocol Television（IPTV）, terminal equipment, network topology, fault location, reliability calculation

摘要： 随着电信IPTV业务的高速发展，承载网规模不断扩大，设备故障运维难度逐渐增大。当设备发生故障时，如何在大规模网络中对故障节点进行快速定位已成为运维方面的重大挑战。目前承载网设备故障告警主要依赖于设备性能日志，误告较多，且无法适应大规模网络故障定位。因此借助Spark等工具，提出一种基于终端数据的异常节点定位方法，结合网络拓扑初步实现阈值告警。并进一步以设备相关性和可靠性为分析基础，将告警过程产生的大量虚假告警进行清洗，提高故障定位精确率。实验结果表明，在承载网故障定位中该方法精确率能达到89%，具有较高实用价值。

关键词: 网络电视（IPTV）, 终端设备, 网络拓扑, 故障定位, 可靠性计算