eDelta: Pinpointing energy deviations in smartphone apps via comparative trace analysis

Abstract

Many smartphone apps can consume an unnecessarily high amount of energy, shortening battery life. Although users can easily notice the undesired fast battery drain, it is almost impossible for them to precisely remember how the abnormal battery drain (ABD) is triggered, making it difficult for developers to fix the problem. Therefore, app developers are in an urgent need for a tool that can provide them helpful information. In this paper, we propose eDelta, a framework that assists developers in pinpointing the APIs with high energy deviation, which usually have a high probability of being relevant to the non-deterministic ABD. Specifically, eDelta performs comparative trace analysis to identify APIs that have significant energy consumption deviation in different user traces. With the information provided by eDelta, developers can substantially reduce the time they spend searching for the ABD root causes. We have prototyped eDelta in Android 4.4 and evaluated it with twenty real-world apps. Our results show that eDelta can effectively pinpoint the APIs with high energy deviation and those APIs indeed cause ABD. Specifically, it reduces, on average, 94.6% of the amount of code that the developers would need to search for ABD root causes.