On the Stability Region of Intermittent Interference Networks

TL;DR

This paper investigates the stable throughput region of a two-pair intermittent interference network, proposing a queue-based protocol, comparing it to capacity, and analyzing message lifetime and complexity trade-offs.

Contribution

It introduces a queue-based transmission protocol for stochastic message arrivals in interference networks and compares stable throughput to capacity, highlighting new requirements for optimal performance.

Findings

Stable throughput region matches capacity when known.

Achieving optimal throughput requires new techniques beyond prior work.

Message lifetime scales with the square root of total communication time.

Abstract

Recent information-theoretic studies have resulted in several interference management (IM) techniques that promise significant capacity improvements over interference avoidance techniques. However, in practice, the stable throughput region is a more relevant metric compared to the capacity region. In this work, we focus on the stable throughput region of a two-pair intermittent interference network with distributed transmitters and propose a queue-based transmission protocol in different regimes to handle the data between queues. In this context, we translate physical-layer IM protocols to accommodate stochastic message arrivals. To evaluate our proposed techniques, we compare the stable throughput region to the capacity region and show, through simulations, that the stable throughput region matches the capacity region when the latter is known. We show that in order to achieve the optimal stable throughput region, new ingredients are needed when compared to prior results. We quantify the trade-off between the encoding/decoding complexity of the proposed scheme (in terms of number of required algebraic operations), and the achievable rates. Finally, we study the lifetime of messages (i.e. the duration from arrival to successful delivery) vis-a-vis the total communication time, and we observe that the average lifetime scales as the square root of the total communication time.