App Parameter Energy Profiling: Optimizing App Energy Drain by Finding Tunable App Parameters
share
In this paper, we observe that modern mobile apps come with a large number of parameters that control the app behavior which indirectly affect the app energy drain, and using incorrect or non-optimal values for such app parameters can lead to app energy drain deficiency or even energy bugs. We argue conventional app energy optimization using an energy profiler which pinpoints energy hotspot code segments in the app source code may be ineffective in detecting such parameter-induced app energy deficiency. We propose app parameter energy profiling which identifies tunable app parameters that can reduce app energy drain without affecting app functions as a potentially more effective solution for debugging such app energy deficiency. We present the design and implementation of Medusa, an app parameter energy profiling framework. Medusa overcomes three key design challenges: how to filter out and narrow down candidate parameters, how to pick alternative parameter values, and how to perform reliable energy drain testing of app versions with mutated parameter values. We demonstrate the effectiveness of Medusa by applying it to a set of Android apps which successfully identifies tunable energy-reducing parameters.
READ FULL TEXT