On the Tradeoff between Energy Harvesting and Caching in Wireless Networks

TL;DR

This paper proposes an online energy-efficient power control scheme for energy harvesting small cell base stations that leverages caching to reduce backhaul usage, balancing energy and storage resources to optimize network performance.

Contribution

It introduces a dynamic programming-based power control method that accounts for stochastic energy arrivals and content popularity, enabling tradeoffs between energy harvesting and caching.

Findings

Caching reduces backhaul usage significantly.

Energy harvesting and cache size can be traded off.

Proposed scheme improves energy efficiency and throughput.

Abstract

Self-powered, energy harvesting small cell base stations (SBS) are expected to be an integral part of next-generation wireless networks. However, due to uncertainties in harvested energy, it is necessary to adopt energy efficient power control schemes to reduce an SBSs' energy consumption and thus ensure quality-of-service (QoS) for users. Such energy-efficient design can also be done via the use of content caching which reduces the usage of the capacity-limited SBS backhaul. of popular content at SBS can also prove beneficial in this regard by reducing the backhaul usage. In this paper, an online energy efficient power control scheme is developed for an energy harvesting SBS equipped with a wireless backhaul and local storage. In our model, energy arrivals are assumed to be Poisson distributed and the popularity distribution of requested content is modeled using Zipf's law. The power control problem is formulated as a (discounted) infinite horizon dynamic programming problem and solved numerically using the value iteration algorithm. Using simulations, we provide valuable insights on the impact of energy harvesting and caching on the energy and sum-throughput performance of the SBS as the network size is varied. Our results also show that the size of cache and energy harvesting equipment at the SBS can be traded off, while still meeting the desired system performance.