Exact Modeling of the Performance of Random Linear Network Coding in Finite-buffer Networks

TL;DR

This paper develops an exact Markov chain-based model to analyze the performance of Random Linear Network Coding in finite-buffer wired erasure networks, capturing buffer dependencies and accurately predicting throughput.

Contribution

It introduces a novel Markov chain model for RLNC in finite-buffer networks, including a reduced-state class for acyclic networks, improving accuracy over classical models.

Findings

Model accurately predicts network throughput with simulations.

Buffer dependencies are effectively captured by the Markov chain approach.

Reduced-state model simplifies analysis for acyclic networks.

Abstract

In this paper, we present an exact model for the analysis of the performance of Random Linear Network Coding (RLNC) in wired erasure networks with finite buffers. In such networks, packets are delayed due to either random link erasures or blocking by full buffers. We assert that because of RLNC, the content of buffers have dependencies which cannot be captured directly using the classical queueing theoretical models. We model the performance of the network using Markov chains by a careful derivation of the buffer occupancy states and their transition rules. We verify by simulations that the proposed framework results in an accurate measure of the network throughput offered by RLNC. Further, we introduce a class of acyclic networks for which the number of state variables is significantly reduced.