On the impact of sampling frequency on software energy measurements

, , , ,
1 Software Cost-effective Change and Evolution Research Lab, Polytechnique Montréal, Montréal, Quebec, Canada
2 Department of Computer Science, The University of Texas at Dallas, Montréal, United States of America
3 MIST Lab, Polytechnique Montréal, Montréal, Quebec, Canada
4 Software Analytics and Technologies Lab, Polytechnique Montréal, Montréal, Quebec, Canada
DOI
10.7287/peerj.preprints.1219v1
Published
Accepted
Subject Areas
Mobile and Ubiquitous Computing, Programming Languages, Software Engineering
Keywords
Performance, Software Energy Consumption, Android
Copyright
© 2015 Saborido et al.
Licence
This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ PrePrints) and either DOI or URL of the article must be cited.
Cite this article
Saborido R, Arnaoudova V, Beltrame G, Khomh F, Antoniol G. 2015. On the impact of sampling frequency on software energy measurements. PeerJ PrePrints 3:e1219v1

Abstract

Energy consumption is a major concern when developing and evolving mobile applications and researchers are investigating ways to reduce energy consumption. We conjecture that these studies are at the border between hardware and software and we must be careful on how the energy consumption is measured. To the best of our knowledge, no previous work investigates how much energy and power consumption is due to high frequency events missed when sampling at low frequencies such as 10 kHz and verified the error at the precision of method level. In this paper, we propose an approach for accurate measurements of the energy consumption of mobile applications. We apply the proposed approach to assess the energy consumption of 21 mobile, closed source, applications and four open source Android applications. We show that by sampling at 10 kHz one may expect a median error of 8%, however, such error may be as high as 50%.

Author Comment

The sampling frequency on software energy measurements has an impact. This document will be a submission to PeerJ for review or some conference in a future.