Comprehensive bottleneck detection via non-linear optimization techniques

Proposes a new traffic analysis method to detect a bottleneck in a large-scale open-type network like the Internet. Unlike conventional queuing theory-based analysis methods, which assume given network configurations and services, the proposed method uses non-linear optimization with an artificial neural network (ANN) using the traffic information collected from an appropriate observation point. We have succeeded in globally estimating the network performance by the detection of a bottleneck, which has not been attained by conventional methods. This paper also suggests the effectiveness of the proposed method with real traffic data.