eDoctor: Automatically Diagnosing Abnormal Battery Drain Issues on Smartphones
Authors:
Xiao Ma, University of Illinois at Urbana-Champaign and University of California, San Diego; Peng Huang and Xinxin Jin, University of California, San Diego; Pei Wang, Peking University; Soyeon Park, Dongcai Shen, Yuanyuan Zhou, Lawrence K. Saul, and Geoffrey M. Voelker, University of California, San Diego
Abstract:
The past few years have witnessed an evolutionary change in the smartphone ecosystem. Smartphones have gone from closed platforms containing only pre-installed applications to open platforms hosting a variety of third-party applications. Unfortunately, this change has also led to a rapid increase in Abnormal Battery Drain (ABD) problems that can be caused by software defects or misconfiguration. Such issues can drain a fully-charged battery within a couple of hours, and can potentially affect a significant number of users.
This paper presents eDoctor, a practical tool that helps regular users troubleshoot abnormal battery drain issues on smartphones. eDoctor leverages the concept of execution phases to capture an app’s time-varying behavior, which can then be used to identify an abnormal app. Based on the result of a diagnosis, eDoctor suggests the most appropriate repair solution to users. To evaluate eDoctor’s effectiveness, we conducted both in-lab experiments and a controlled user study with 31 participants and 17 real-world ABD issues together with 4 injected issues in 19 apps. The experimental results show that eDoctor can successfully diagnose 47 out of the 50 use cases while imposing no more than 1.5% of power overhead.
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.
BibTeX
@inproceedings {180295,
author = {Xiao Ma and Peng Huang and Xinxin Jin and Pei Wang and Soyeon Park and Dongcai Shen and Yuanyuan Zhou and Lawrence K. Saul and Geoffrey M. Voelker},
title = {{eDoctor}: Automatically Diagnosing Abnormal Battery Drain Issues on Smartphones},
booktitle = {10th USENIX Symposium on Networked Systems Design and Implementation (NSDI 13)},
year = {2013},
isbn = {978-1-931971-00-3},
address = {Lombard, IL},
pages = {57--70},
url = {https://www.usenix.org/conference/nsdi13/technical-sessions/presentation/ma},
publisher = {USENIX Association},
month = apr
}
Smartphones have evolved from closed platforms that run a few pre-installed applications to open systems that run a variety of first-party and third-party programs. Modern smartphones also contain a variety of sensors that applications can use to capture audio and video, monitor acceleration, and detect location. Unfortunately, as the smartphone ecosystem has grown richer, smartphones have become more vulnerable to abnormal battery drain (ABD) in which application bugs, bad configuration state, or unexpected environmental conditions cause rapid battery draining. For example, a faulty application may continuously poll the GPS unit, keeping the unit powered up even though the user is location is not changing rapidly. As another example, a well-intentioned user may misconfigure a weather application, causing it to frequently pull weather updates across the network, leading to a dramatic increase in network-related energy usage.
eDoctor is a new system for diagnosing ABD in smartphones. eDoctor monitors each application is usage of resources like the CPU, the wireless radio, and the GPS unit. eDoctor breaks this time series into a set of intervals, and identifies common and uncommon phases, with each phase representing a particular pattern of resource usage (e.g.,high CPU utilization but low network usage). eDoctor also measures an application is energy consumption, as well as system events like application installation and configuration. To diagnose ABD, eDoctor combines these three sources of information (phase resource utilization, phase energy usage, and system-wide event logs). eDoctor first identifies abnormal, energy-heavy phases, and then looksfor system events that preceded the onset of the abnormal phases; these events are likely causes for the ABD.
To evaluate eDoctor, the authors performed a controlled user study. The authors found 19 applications with 21 ABD problems (17 problems were found in the wild, and 4 were synthetically injected). The authors distributed these applications to 31 smartphone users. The users interacted with known-good versions of the applications for at least 5 days. Users then converted the applications to known-ABD versions, interacted with them until ABD was apparent, and then used eDoctor to diagnose the issue. eDoctor accurately diagnosed 94% of the ABD root causes. eDoctor added little overhead, requiring less than 1.5% additional power and only 22.4 MB of storage space.
In general, reviewers thought that this paper attacked an important problem. The authors clearly built a real tool with practical usefulness, and the evaluation suggested that eDoctor performed well under realistic conditions.
connect with us