Fault resolution and alarm correlation in high-speed networks using database mining techniques

A telecommunications company must be able to effectively manage its network. Increasing system size, complexity and bandwidth are threatening to overload current management systems so it comes as no surprise that network management is a high priority for leading telecommunications providers. Central to network management is the monitoring of the performance of the network and the handling of large volumes of alarm messages to resolve or avoid faults. Although many systems have been proposed that correlate alarms, the more fundamental question still remains of what to correlate and how to recognise it. For large telecommunications networks, complexity means that explicit symptoms for all possible occurrences cannot easily be determined. We examine the role that data mining can play in the analysing and generalisation of the sizeable volumes of network performance information that are collected every day. We describe its potential use as a tool for network fault recognition, discovery and categorisation, verifying our approach using alarm data from an SDH transmission simulation tool