Efficient Queue-based CSMA with Collisions

TL;DR

This paper extends the analysis of efficient CSMA protocols to scenarios with imperfect or delayed carrier sense information, demonstrating throughput optimality and providing foundational results for collision-aware rate control in wireless networks.

Contribution

It adapts a known CSMA algorithm for delayed carrier sense information and establishes its throughput optimality through analysis of a non-reversible Markov chain.

Findings

Throughput optimality of the adapted CSMA algorithm.

Analysis of stationary distribution and mixing time of the Markov chain.

Framework for collision-aware rate control in wireless networks.

Abstract

Recently there has been considerable interest in the design of efficient carrier sense multiple access(CSMA) protocol for wireless network. The basic assumption underlying recent results is availability of perfect carrier sense information. This allows for design of continuous time algorithm under which collisions are avoided. The primary purpose of this note is to show how these results can be extended in the case when carrier sense information may not be perfect, or equivalently delayed. Specifically, an adaptation of algorithm in Rajagopalan, Shah, Shin (2009) is presented here for time slotted setup with carrier sense information available only at the end of the time slot. To establish its throughput optimality, in additon to method developed in Rajagopalan, Shah, Shin (2009), understanding properties of stationary distribution of a certain non-reversible Markov chain as well as bound on its mixing time is essential. This note presents these key results. A longer version of this note will provide detailed account of how this gets incorporated with methods of Rajagopalan, Shah, Shin (2009) to provide the positive recurrence of underlying network Markov process. In addition, these results will help design optimal rate control in conjunction with CSMA in presence of collision building upon the method of Jiang, Shah, Shin, Walrand (2009).