Analysis of Non-Persistent CSMA Protocols with Exponential Backoff Scheduling

TL;DR

This paper analyzes the performance of Non-persistent CSMA/CA protocols with exponential backoff, providing a comprehensive analysis of throughput, delay, and stability regions for different backoff schemes in multi-access networks.

Contribution

It offers the first complete analysis of throughput and delay distributions for K-Exponential Backoff in non-persistent CSMA protocols, including stability conditions for large networks.

Findings

Geometric Retransmission becomes unstable as node count increases.

Exponential Backoff maintains stability for infinite network sizes.

Delay variance can be unbounded due to capture effects.

Abstract

This paper studies the performance of Non-persistent CSMA/CA protocols with K-Exponential Backoff scheduling algorithms. A multi-queue single-server system is proposed to model multiple access networks. The input buffer of each access node is modeled as a Geo/G/1 queue, and the service time distribution of head-of-line packets is derived from the Markov chain of underlying scheduling algorithm. The main results include the complete analysis of the throughput and delay distribution, from which we obtained stable regions with respect to the throughput and bounded mean delay of the Geometric Retransmission and Exponential Backoff schemes. We show that the throughput stable region of Geometric Retransmission will vanish as the number of nodes n \rightarrow \infty; thus, it is inherently unstable for large n. In contrast to Geometric Retransmission, the throughput stable region of Exponential Backoff can be obtained for an infinite population. We found that the bounded mean delay region of Geometric Retransmission remains the same as its throughput stable region. Besides, the variance of service time of Exponential Backoff can be unbounded due to the capture effect; thus, its bounded delay region is only a sub-set of its throughput stable region. Analytical results presented in this paper are all verified by simulation.