Throughput Scaling Laws for Wireless Networks with Fading Channels

TL;DR

This paper analyzes the throughput of large wireless networks with fading channels, proposing decentralized and centralized power allocation schemes that improve per-link rates and overall network throughput.

Contribution

It introduces a decentralized on-off power scheme that achieves a logarithmic sum-rate under Rayleigh fading and a centralized scheme that enhances per-link rates for large networks.

Findings

Decentralized scheme achieves sum-rate of order log n under Rayleigh fading.

Allowing a small reduction in sum-rate yields constant per-link rates.

Centralized scheme outperforms decentralized in per-link rate for the same number of active links.

Abstract

A network of $n$ wireless communication links is considered. Fading is assumed to be the dominant factor affecting the strength of the channels between nodes. The objective is to analyze the achievable throughput of the network when power allocation is allowed. By proposing a decentralized on-off power allocation strategy, a lower bound on the achievable throughput is obtained for a general fading model. In particular, under Rayleigh fading conditions the achieved sum-rate is of order $\log n$, which is, by a constant factor, larger than what is obtained with a centralized scheme in the work of Gowaikar et al. Similar to most of previous works on large networks, the proposed scheme assigns a vanishingly small rate for each link. However, it is shown that by allowing the sum-rate to decrease by a factor $\alpha<1$, this scheme is capable of providing non-zero rate-per-links of order $\Theta(1)$. To obtain larger non-zero rate-per-links, the proposed scheme is modified to a centralized version. It turns out that for the same number of active links the centralized scheme achieves a much larger rate-per-link. Moreover, at large values of rate-per-link, it achieves a sum-rate close to $\log n$, i.e., the maximum achieved by the decentralized scheme.