Offline Trace Synchronization for Smartphone Energy Profiling

Energy profiling is a means to effectively understand the power behavior of smartphone applications. However, no tool that effectively combines portability, accuracy and automation has been proposed yet. In this thesis a new approach is proposed, which sets the basis for such a tool by solving a trace synchronization problem. TRAM (TRAce Merger) uses the event trace from a mobile device and the power trace from a measurement device. It synchronizes them offline, maps the mobile device events trace on the corresponding trace from the power measurement tool, and attributes energy consumption to smartphone functionalities. The event and power traces traces from these two systems contain timestamps based on the corresponding local clocks. Taking the aforementioned clocks inaccuracy and deviation into account, we synchronize these traces to extract useful information from their contents. We achieved energy consumption attribution to smartphone functionalities by solving this synchronization problem with milliseconds accuracy.