Towards more intelligent network management: service-oriented proactive fault management using KDD techniques

In this paper, we present a knowledge discovery system, Data Fusion Supported Frequent Episodes Discovery (DFSPED), for service-oriented proactive fault management in communication networks. The concept of network service topology is proposed first, and context-free grammars and network service dependence graph are used to model the service-oriented proactive fault management. Secondly, with the approaches of data fusion, two catalogues of important fault-pertinent data sources, i.e., service-oriented trouble-reporting data and real time alarm messages, are fused by temporal and spatial associations. KDD techniques are applied to perform incremental mining in the fused sequential events, finding frequent episodes among faults, alarms and service. Fast restoration of key service interruption, fault prediction and preventive maintenance benefit a lot from the discovered episodes