Analyzing Impact of Different Inter-Arrival Time Distributions on the Performance of IoT Systems
Prakash Datt Bhatt, Shachi Sharma · 2022
Poisson arrival-based analytical traffic models are found to be inappropriate for IoT networks. Many empirical studies confirm that beta, gamma, generalized exponential, generalized Pareto, lognormal and Weibull distributions characterize the inter-arrival time of different types of IoT traffic. It then becomes imperative to examine the impact of various arrival patterns on the quality of service measures of IoT systems. The paper presents GE/M/1 model where inter-arrival time follows generalized exponential distribution and service time is exponentially distributed. The key finding of the analysis of GE/M/1 model is that the mean queue length as well as mean delay remain lower than Poisson arrivals-based M/M/1 when shape parameter is greater than scale and vice-versa. Since, traffic models of generalized Pareto and Weibull distributions are analytically intractable, a detailed comparative performance study of various types of arrival patterns with exponential and deterministic service times have been performed using Monte-Carlo simulation. The results reveal that the parameters of inter-arrival time distributions play key role in determining the performance of IoT systems. These findings are useful for designing efficient congestion control algorithms for IoT networks.