An App Performance Optimization Advisor for Mobile Device App Marketplaces

TL;DR

This paper introduces APOA, a recommendation system that helps users and developers compare mobile apps based on performance metrics, addressing the lack of performance information in app marketplaces.

Contribution

The paper presents APOA, a novel optimization-based recommendation system for app marketplaces that considers performance metrics to assist in app selection.

Findings

APOA effectively compares apps using power, CPU, memory, and network metrics.

APOA generates optimal app sets tailored to user scenarios.

Case study with 140 Android apps demonstrates APOA's benefits.

Abstract

On mobile phones, users and developers use apps official marketplaces serving as repositories of apps. The Google Play Store and Apple Store are the official marketplaces of Android and Apple products which offer more than a million apps. Although both repositories offer description of apps, information concerning performance is not available. Due to the constrained hardware of mobile devices, users and developers have to meticulously manage the resources available and they should be given access to performance information about apps. Even if this information was available, the selection of apps would still depend on user preferences and it would require a huge cognitive effort to make optimal decisions. Considering this fact we propose APOA, a recommendation system which can be implemented in any marketplace for helping users and developers to compare apps in terms of performance. APOA uses as input metric values of apps and a set of metrics to optimize. It solves an optimization problem and it generates optimal sets of apps for different user's context. We show how APOA works over an Android case study. Out of 140 apps, we define typical usage scenarios and we collect measurements of power, CPU, memory, and network usages to demonstrate the benefit of using APOA.