IoT system evaluation methods for very bursty traffic with contention based access

Abstract

One very popular IoT system is characterized by the very bursty traffic in which the transmission occurrence can be as lower as just 200 milliseconds in multiple hours. The traditional Monte Carlo behavioral simulation is very time-consuming for a usual IoT system where the traffic burst is significantly smaller than the hibernation time and the number of terminals is big. Alternatives are analytical approaches using random probability and stochastic process theory, and there have been numerous papers on Slotted ALOHA (S-ALOHA) system. Unfortunately almost all the results obtained so far are based on full buffer traffic model so there is not an efficient approach to evaluate the system performance of an IoT system supporting massive number of terminals with very bursty traffic patterns. Two analytic methodologies are proposed and studied in this paper for very bursty traffic pattern in a contention based access system employing S-ALOHA access with Binary Exponential Back-off (BEB), and a set of closed-form formulae are derived. The first method is based on the probability modeling and the second method is a Markov chain based 2-D analytical model. The fundamental parameter in both methods is the expression for packet transmission probability, and the close-form expressions are derived for both methods. It was verified that the two expressions agree with each other very well. Based on the packet transmission probability, the close form expressions for other system performance indicators are also derived, which includes packet loss rate, transmission time, transmission delay, re-transmission times, collision possibility, and channel utilization efficiency. The proposed methods were verified to match real system performance very well, and they can be the efficient and accurate analytical tools for IoT system performance evaluation and optimization supporting very bursty traffic.