A Method of Large - Scale Log Pattern Mining

With the development of the telecommunication network, more and more devices are used in the network, which has been a burden for the network operation and maintenance. At the same time, network devices generate large amounts of log data every day, recording the activities of each device in detail. As a result, the log can reflect the performance of network state, and sometimes, we can predict the occurrence of network failure based on the log. However, since the log has such features: big volume, multi-source heterogeneous and difficult to understand, people have not reasonably used it to analyze and predict network failure. Therefore, we propose a method for structuring a large number of device logs in the short term, and use the data generated from a real communication device network to verify the effect. Besides, we compare our method with the traditional log parsers, such as regular expressions, LogSig, etc. to demonstrate the efficient processing performance and accurate pattern extraction analysis for massive network device logs.