| With the rapid development of the Communication Network, the scale of the network is growing larger and network technology has become increasingly complex. China Mobile, for real-time monitoring has to treatment different manufacturers (Ericsson, Huawei, ZTE, etc.) and variety alarm information from different network equipment. Currently Mobile monitors alarm information depending on the management system provided by different vendors, which is of great inconvenience to unified management device and analysis alarm data. So to build a unified network fault management system for effectively maintaining and ensuring the quality of the network become the needs of mobile and telecom operators.This paper in-depth analysis functions and key technologies of network equipment fault management which cooperates with Mobile and proposes the solutions of fault management based on Telnet protocol. It discusses the real-time and efficiency problem with network 10 model when connecting a large number of network equipment to connect alarm information. It also talks over alarm format analytical for verity types of devices. In order to find the relation among data, the paper researches the alarm correlation model.With the comparison of five network 10 models, the paper adopts IOCP which has good scalability. With this 10 model, it achieves fast collecting alarm data. Through an unified analysis for different data formats from different types of devices to achieve a real-time monitoring, data storage and query. Research on the data mining models based on association rules obtains association rules with simulating.Finally, the system performance is tested and evaluated, in the case of more than 70 concurrently connected network equipments working, and the performance is that, the gathering speed is over 10 records per second. In half of one year, more than 200 millions alarm records are collected. All the functions meet the requirements of China Mobile Company at Dalian, which proves that the system owns perfect stability and reliability. After the measurements, the analytical model to obtain the rules of alarm association is proved of some practical value in the alarm analysis of communication network. |