An interest in traffic analysis has been increasing to evaluate the internet performance and stability between ISPs(Internet service Providers) In general, two parameters such as delay and packet loss are used for performance analysis Existing studies...
An interest in traffic analysis has been increasing to evaluate the internet performance and stability between ISPs(Internet service Providers) In general, two parameters such as delay and packet loss are used for performance analysis Existing studies only discuss traffic measurement and simple performance test to check if measured values exceed their thresholds In this paper, we present a BP(Back Propagation) based evaluation scheme for performance estimation. The BP is a neural network, which can estimate the target performance continuously applying supervised learning patterns We have collected real data by measuring delay and loss over a commercial ISP line last two months and performed the BP computation using these data Then we show that the estimation mapping formula obtained between loss and delay can be used for performance evaluation metrics.