Temporal alarm pattern discovery in mobile telecommunication networks based on binary series analysis

摘要

Highly-advanced systems, such as mobile telecommunication networks, characterized by increased complexity, make maintenance routines difficult. Amount of data to be analyzed in a short time during fault diagnosis of the mobile telecommunication networks strongly justifies the need to automate alarm correlation and root cause analysis. A major challenge in the establishment of alarm correlation is to determine how to reflect the alarm flow inertia. Thus, adequate temporal alarm pattern discovery methods should be used in fault diagnosis for correlation-related purposes. Automatic temporal alarm pattern discovery allows fast generation of root cause analysis hypotheses and supports effective troubleshooting of network problems. The process for fault propagation throughout the network is manifested by the time lag between the root-cause alarm and potentially linked symptoms, as well as weakening correlation strength with time. The paper presents a novel method for alarm correlation analysis in mobile telecommunication networks, based on binary series analysis. The method allows for discovery of causal relationship between alarms with dynamic alarm correlation window size estimation.