Traffic modeling for aggregated periodic IoT data

The Internet of Things (IoT) is emerging in the telecommunication sector, and will bring a very large number of devices that connect to the Internet in the near future. The expected growth in such IoT nodes necessitates appropriate traffic models in order to evaluate their impact on different aspects of networking, e.g., on signaling load in the networks, or on processing load of the data in a cloud. In this paper we analyze the characteristics of aggregated periodic IoT data based on related work, and compare them with a Poisson process as approximation for the superposed traffic as assumed in standardization. Such an approximation is crucial in order to investigate the scalability of an IoT network, as it may be impossible in practice to measure or to simulate large-scale IoT deployments. The accuracy and applicability of the Poisson process is investigated for the use case 'IoT cloud'. The results show that the Poisson process may induce large errors depending on the performance metric of interest. This error must be considered by standardization and requires more sophisticated traffic models. As key contributions, we provide realistic traffic models for periodic IoT data, introduce performance metrics for quantifying the bias, and derive reference values as to when the Poisson process can be assumed for aggregated periodic IoT data.