| The increasing role of communication networks in today's society results in a demand for higher levels of network availability and reliability. As our dependency on networks increases, faults and downtime become very costly. At the same time, fault management is becoming more difficult due to the dynamics and heterogeneity of the network. Current network management systems rely heavily on the availability of network expertise. These systems must be customized to a particular network, and as such cannot be easily generalized from one network to another.; To improve fault management in the high-speed communication networks of today, we propose an intelligent monitoring system using adaptive learning machines. The system continually learns the normal behavior of the network and detects deviations from the norm. Within the monitoring system, the measurements are segmented, and features extracted from the segments are used to describe the normal behavior of the measurement variables in terms of a probability distribution. The likelihood of each measurement being abnormal is estimated based on the normal behavior distribution. This information is combined in the structure of a Bayesian network. Novel aspects of the system include the ability to correlate the measurements in space and time, and the inclusion of new features.; By detecting anomalies instead of specific faults, we are able to detect problematic behavior before a fault actually occurs. This can potentially enable actions to be taken to avoid a serious problem. Since the system requires minimal a priori knowledge about a specific network, it can also be generalized to other networks. We demonstrate the system on network data collected from the RPI CS network. Experimental results show that this method can detect abnormal behavior before a fault actually occurs. |
