Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Measuring the Energy of Smartphone Communications in the Edge-Cloud Continuum: Approaches, Challenges, and a Case Study

Chiara Caiazza, Valerio Luconi, Alessio Vecchio

TL;DR

Measuring smartphone energy consumption during edge-cloud communications addresses how resource placement affects energy use on constrained devices. The paper surveys analytical models, network simulations, and software/hardware monitors for estimating communication energy and reports a case study comparing edge, cloud, and far-cloud deployments, showing edge can reduce energy by substantial margins. A simple energy expression is used, $E_T = P_{sleep} T_{sleep} + P_{awake} T_{awake}$, with edge energy observed to be $54-40\%$ of cloud energy for 2–12 MB resources, and results depend on payload size and network conditions. The work highlights practical measurement challenges and suggests AI-assisted orchestration as a future direction to optimize energy in the edge-cloud continuum.

Abstract

As computational resources are placed at different points in the edge-cloud continuum, not only the responsiveness on the client side is affected, but also the energy spent during communications. We summarize the main approaches used to estimate the energy consumption of smartphones and the main difficulties that are typically encountered. A case study then shows how such approaches can be put into practice. Results show that the edge is favorable in terms of energy consumption, compared to more remote locations.

Measuring the Energy of Smartphone Communications in the Edge-Cloud Continuum: Approaches, Challenges, and a Case Study

TL;DR

Measuring smartphone energy consumption during edge-cloud communications addresses how resource placement affects energy use on constrained devices. The paper surveys analytical models, network simulations, and software/hardware monitors for estimating communication energy and reports a case study comparing edge, cloud, and far-cloud deployments, showing edge can reduce energy by substantial margins. A simple energy expression is used, , with edge energy observed to be of cloud energy for 2–12 MB resources, and results depend on payload size and network conditions. The work highlights practical measurement challenges and suggests AI-assisted orchestration as a future direction to optimize energy in the edge-cloud continuum.

Abstract

As computational resources are placed at different points in the edge-cloud continuum, not only the responsiveness on the client side is affected, but also the energy spent during communications. We summarize the main approaches used to estimate the energy consumption of smartphones and the main difficulties that are typically encountered. A case study then shows how such approaches can be put into practice. Results show that the edge is favorable in terms of energy consumption, compared to more remote locations.

Paper Structure

This paper contains 13 sections, 1 equation, 3 figures.

Table of Contents

  1. Introduction and Background
  2. Approaches
  3. Challenges
  4. Analytical and Simulation Approaches
  5. Software and Hardware Approaches
  6. Software monitors
  7. Hardware monitors
  8. Case Study: The Impact of Latency in the Edge-Cloud Continuum
  9. Analytical Approach
  10. Software Monitor
  11. Hardware Monitor
  12. Discussion and Future Directions
  13. Acknowledgment

Figures (3)

  • Figure 1: The finite state machine of an LTE module.
  • Figure 2: Absorbed current for the three edge/cloud configurations, measured with a software monitor. For each payload size, we selected the minimum value of mean absorbed current computed for slots of one hour each. In this case, the cloud and the far cloud servers are emulated by artificially adding 100 ms and 200 ms to the path with tc.
  • Figure 3: Power consumption for the three edge/cloud configurations, measured with a hardware power monitor. For each resource size, we select the minimum value of mean power computed for slots of 10 minutes each. Differently from the software monitor use case, here, edge, cloud, and far cloud servers are real-world servers from the GCS infrastructure. In particular, they are located in Italy, South Carolina, and Australia.