This conference paper takes the view that by adding declarative annotations and middleware components to an Android software development platform, energy savings can be achieved in embedded deployed applications. The annotation highlights power-hungry code segments whose execution gets deferred. The envisaged application is mostly in mobile health platforms. The paper provides an abstract model of the approach based on time-lapse automata, nothing particularly new. Policy examples, expressed as Java annotations, are described in the context of set power management systems (especially the CitiSense case).
The paper is rather theoretical and largely devoted to the specification of the annotation primitives. The runtime is made up of a client library and a lightweight middleware service communicating by remote procedure calls (RPCs). The related overhead is evaluated in the context of one case. The relative contributions of the preexisting power management system and the added annotation are not clear. This approach may serve as a first cut in energy-critical applications, but does not allow for fine-tuning with regards to the power management system itself or the application code.
For energy critical performance, optimization at the compiler level with relevant energy consumption attributes is much more efficient because of the scheduling that takes place in the compiler.