A Survey on Event Mining for ICT Network Infrastructure Management

Release Date:2016-05-12 Author:LIU Zheng, LI Tao, and WANG Junchang Click:

[Abstract] Managing large⁃scale complex network infrastructures is challenging due to the huge number of heterogeneous network elements. The goal of this survey is to provide an overview of event mining techniques applied in the network management domain. Event mining includes a series of techniques for automatically and effectively discovering valuable knowledge from historical event/log data. We present three research challenges (i.e., event generation, root cause analysis, and failure prediction) for event mining in network management and introduce the corresponding solutions. Event generation (i.e., converting messages in log files into structured events) is the first step in many event mining applications. Automatic root cause analysis can locate the faulty elements/components without the help of experienced domain experts. Failure prediction in proactive fault management improves network reliability. The representative studies to address the three aforementioned challenges are reviewed and their main ideas are summarized in the survey. In addition, our survey shows that using event mining techniques can improve the network management efficiency and reduce the management cost.

[Keywords] event mining; failure prediction; log analysis; network infrastructure management; root cause analysis

Download: PDF