Learning the parameters of periodic traffic based on network measurements

Abstract

The configuration of real-time networks is one of the most challenging demands of the Real-Time Internet-of-Things trend, where the network has to be deterministic and yet flexible enough to adapt to changes through its life-cycle. To achieve this we have outlined an approach that learns the necessary configuration parameters from network measurements, that way providing a continuous configuration service for the network. First, the network is monitored to obtain traffic measurements. Then traffic parameters are derived from those measurements. Finally, a new time-triggered schedule is produced with which the network will be reconfigured. In this paper we propose an analysis based on measurements to obtain the specific traffic parameters and we evaluate it through network simulations. The results show that the configuration parameters can be learned from the measurements with enough accuracy and that those measurements can be easily obtained through network monitoring.