Comparison of M2M Traffic Models Against Real World Data Sets

Abstract

Machine-To-Machine (M2M) traffic is expected to significantly increase in future wireless networks. In order to study the effect of this type of traffic in current and future networks, there is the need for efficient and effective traffic models. The literature offers different models, but there is not general consensus on which of them can better represent realistic M2M traffic sources. In this paper, we contribute to shed light on this aspect in two ways: first, we analyze some real-world data traces, provided by one of the biggest M2M operators in Europe, to have a better idea of the characteristics of realistic traffic patterns; second, we compare the capabilities of three popular and flexible M2M source models proposed in the literature to reproduce the empirical data patterns, and we suggest some possible improvements. The analysis reveals that the traffic patterns generated by the considered M2M services have strong deterministic components, which require to increase the determinism of the source models to improve their accuracy.