| （英） |
With the diversification of applications, network control based on QoE, which can uniformly evaluate application quality, has been attracting attention. In our previous study, we proposed a method for estimating and utilizing the QoE of video in a software-defined network (SDN). However, SDNs inevitably suffer from measurement errors in packet loss due to timing differences in the acquisition of statistical information, and the accuracy of QoE estimation for video, for which packet loss is an important metric, is significantly degraded. Therefore, in this study, we first conducted a comprehensive survey of factors related to the occurrence of measurement errors, and then By using the identified factors as machine learning features, we aim to enable real-time measurement error correction during QoE estimation. We show that there is a significant relationship between jitter in the network between SDN controllers and switches and the occurrence of measurement error, and create a learner that determines measurement error based on this relationship. The results of the study show that jitter in one direction from the SDN controller to the switch is correlated with measurement error, and we propose a method to estimate this one-way jitter. We also show that it is possible to determine the measurement error by using the one-way jitter obtained from the experiments as a feature. |