Performance Analysis of Markov Modulated 1-Persistent CSMA/CA Protocols with Exponential Backoff Scheduling

This paper proposes a Markovian model of 1-persistent CSMA/CA protocols with K-Exponential Backoff scheduling algorithms. The input buffer of each access node is modeled as a Geo/G/1 queue, and the service time distribution of each individual head-of-line packet is derived from the Markov chain of the underlying scheduling algorithm. From the queuing model, we derive the characteristic equation of network throughput and obtain the stable throughput and bounded delay regions with respect to the retransmission factor. Our results show that the stable throughput region of the exponential backoff scheme exists even for an infinite population. Moreover, we find that the bounded delay region of exponential backoff is only a sub-set of its stable throughput region due to the large variance of the service time of input packets caused by the capture effect. All analytical results presented in this paper are verified by simulations.

💡 Research Summary

The paper presents a rigorous analytical framework for evaluating the performance of 1‑persistent CSMA/CA protocols when combined with a K‑stage exponential backoff scheduling algorithm. Each access node is modeled as a Geo/G/1 queue: packet arrivals follow a geometric distribution, while the service process is governed by the backoff and retransmission dynamics of the underlying MAC protocol. The authors first construct a discrete‑time Markov chain whose states capture both the current backoff stage (i = 0,…,K‑1) and whether the most recent transmission attempt resulted in a collision. Transition probabilities are expressed in terms of a retransmission factor ρ and the conditional collision probability, which itself depends on the number of contending nodes N and the instantaneous set of transmitting stations. By solving the steady‑state balance equations, the stationary distribution π(i,c) is obtained, from which the probability mass function of the head‑of‑line (HoL) packet service time T is derived. Crucially, the first two moments E