Characterizing Mobile Open APIs in smartphone apps

Mobile applications used in smartphones are increasingly using Open APIs, and the trend is likely to continue in the foreseeable future. However, the performance of the Open APIs integrated in smartphone apps (Mobile Open APIs) remains hidden from app developers and app users because of the lack of a method to isolate the Open API calls from the whole app execution process. In this paper, we present the very first effort on characterizing Mobile Open APIs and analyzing their performance in terms of four metrics: response latency, network traffic, energy consumption, and CPU usage. We first develop APIExtractor (APIX), a software tool to extract the Mobile Open API calls as fine-grained as in the function level from Android app files (.apk). Then the popularity of the Open API functions were ranked by running APIX on 200 top popular apps downloaded from the Android app store. We then perform in-depth case studies on the the top 17 most popular Mobile Open APIs, by wrapping each of them in a specifically designed app (called APISymphone) and test the apps on both Wi-Fi and cellular network. Furthermore, we conduct a global scale measurements of the Mobile Open APIs by using Amazon Elastic Computing service. Our comprehensive measurement-based results provides very intriguing as well as interesting insights to the performance characteristics of Mobile APIs.