Delay-Sensitive Communication over Fading Channel: Queueing Behavior and Code Parameter Selection

TL;DR

This paper analyzes the queueing performance of delay-sensitive wireless communication systems using coded data over fading channels with memory, providing criteria for optimal code parameter selection based on Markov models and numerical simulations.

Contribution

It introduces a new Markov-based analytical framework for joint queueing and coding analysis over finite-state channels with memory, aiding optimal code design.

Findings

Bounds on queue overflow probabilities derived

Optimal code parameters identified for delay-sensitive scenarios

Numerical simulations validate the analytical model

Abstract

This article examines the queueing performance of communication systems that transmit encoded data over unreliable channels. A fading formulation suitable for wireless environments is considered where errors are caused by a discrete channel with correlated behavior over time. Random codes and BCH codes are employed as means to study the relationship between code-rate selection and the queueing performance of point-to-point data links. For carefully selected channel models and arrival processes, a tractable Markov structure composed of queue length and channel state is identified. This facilitates the analysis of the stationary behavior of the system, leading to evaluation criteria such as bounds on the probability of the queue exceeding a threshold. Specifically, this article focuses on system models with scalable arrival profiles, which are based on Poisson processes, and finite-state channels with memory. These assumptions permit the rigorous comparison of system performance for codes with arbitrary block lengths and code rates. Based on the resulting characterizations, it is possible to select the best code parameters for delay-sensitive applications over various channels. The methodology introduced herein offers a new perspective on the joint queueing-coding analysis of finitestate channels with memory, and it is supported by numerical simulations.