Flexible Installability of Android Apps with App-level Virtualization based Decomposition

TL;DR

This paper introduces AppStarscream, a system that decomposes Android apps into feature-specific components using app-level virtualization, reducing resource consumption and improving installability based on empirical user feature usage data.

Contribution

It presents a novel app decomposition approach with virtualization to enable selective feature installation, addressing resource waste and user preferences.

Findings

Empirical study shows users only use a small subset of app features.

Prototype system demonstrates efficiency and practicality.

Decomposition reduces resource consumption during app installation.

Abstract

With the popularity of smartphones, users are heavily dependent on mobile applications for daily work and entertainments. However, mobile apps are becoming more and more complicated with more features and increasing size, part of which may be redundant to users. Due to the limitation of current installation mechanism, users have to download full-size applications instead of enjoy only the wanted features. Such full-size apps may consume more resources, including CPU, memory, and energy, which may hurt users' enthusiasm for further installation. We first conduct an empirical study to characterize used features when users interact with mobile applications, and find that users only consume a small set features of target apps. To address this problem, we present AppStarscream, which offers to decompose and run Android apps with app-level virtualization. We have implemented a prototype system and evaluated it with real-world apps showing that AppStarscream is efficient and practical.