Investigating the energy use of smart phones
More efficient use of energy would mean your smart phone could be smaller, cheaper or simply recharged less frequently. This talk will use the efficiency of sending data over a wireless network as an example. Our experimental setup embeds synchronization information in the energy trace of the phone to allow us to overlay information from various capture sources. I will explain this process and show the results so far. We have found numerous factors affecting the energy consumption of this example and we are now looking to develop probabilistic models which we can compose together to capture their interactions. The benefit of any optimizations derived from our controlled experiments will depend on real-world usage of phone handsets. In the second part of the talk I will describe our efforts using DeviceAnalyzer to build a publicly available data set of handset usage. Using the data we have collected over a 2 year period from over 10,000 users I will look at what determining factors of energy consumption we can measure or infer and highlight the areas for future work.