Events Calendar

PhD Defense Ding Li

Speaker: Ding Li, PhD Candidate

Talk Title: Energy Optimization of Mobile Applications

Abstract: Energy is a critical resource for mobile devices. Many techniques have been proposed to optimize the energy consumption of mobile devices at the hardware and system levels. However, only optimizations at the hardware and system level are insufficient. Poorly designed applications can still waste the energy of mobile devices even with fully optimized hardware and system support. In my dissertation work, I proposed multiple techniques to help developers to create energy efficient apps. Particularly, my dissertation work addresses three problems in creating energy efficient apps. The first problem in my dissertation is "where is energy consumed." Modern mobile apps are very complex. They may contain more than 500,000 lines of code. Thus, it is important to know which part of the code consumes more energy. To address this problem, I developed a source line level energy measurement technique that can report the energy consumption of mobile apps with a very fine granularity. My technique achieved 91% accuracy during the measurement. The second problem in my dissertation is "what to optimize." Modern mobile apps may use different libraries and invoke thousands of APIs. It is also important to know what kind of libraries and APIs can consume more energy. To address this problem, I conducted an empirical study with 405 Android market apps about how these Android apps consume energy. In this study, I evaluated ten research questions that have motivated my following energy optimization techniques. The third problem is "how to optimize." After knowing where is energy consumed and what to optimize, it is also important to design effective techniques to optimize the energy consumption of mobile apps. To address this problem, I developed two automated techniques. The first one can automatically optimize the display energy for mobile web apps and the second one can optimize HTTP energy for Android apps. My display energy optimization technique reduced the energy by 25% and my HTTP energy optimization technique achieved 15% energy savings. Besides the energy optimization techniques, I also improved the flexibility, accuracy, and efficiency of the string analysis technique, which is very important to my optimization techniques. In summary, my techniques and the empirical evaluation show that program analysis techniques can help developers to understand how energy is consumed in mobile apps and can also help to optimize the energy consumption of mobile apps.

Host: Ding Li

Location: 213

Audiences: Everyone Is Invited

Contact: Ryan Rozan