Periodic Transmissions in Random Access Networks: Stressed Period and Delay

Most IoT systems use random access protocols for wireless communication. This paper considers an IoT node that generates periodic traffic to be delivered to a destination over a random access network; each packet of the node is expected to be delivered before its deadline. We say that the node is in a stressed period if, within a time interval, its successive packets miss their deadlines. In many systems, the worst-case performance is significantly affected by stressed periods. Characterizing the stochastic properties of stressed periods is thus of fundamental importance. In this paper, we use a fluid flow model to approximate the evolution of the buffer occupancy (i.e., backlog) at the transmitting node. We derive a relationship between buffer occupancy and delay and formally define a stressed period via a time interval in which the buffer occupancy exceeds a certain threshold. With this model, we analyze the dynamics of the buffer occupancy evolution and obtain the probability distributions of stressed period duration and delay. Real network experiments show that our model can well approximate the distributions of stressed period duration and delay in practical WiFi networks. The theoretical results of this paper can be used to analyze the robustness and worst-case performance of IoT monitoring and control systems built on random access networks.

Automated summary

This paper investigates the stochastic properties of stressed periods in IoT systems using random access protocols for wireless communication. A fluid flow model is used to approximate the evolution of buffer occupancy at the transmitting node, and a relationship between buffer occupancy and delay is derived. Stressed periods are formally defined as time intervals where the buffer occupancy exceeds a certain threshold, and the probability distributions of stressed period duration and delay are obtained. Real network experiments validate the accuracy of the proposed model and its applicability in analyzing the worst-case performance of IoT systems.