Tech Report CS-11-03

The Case for Device Power States

Marcelo Martins and Rodrigo Fonseca

October 2011

Abstract:

Understanding power draw in mobile devices has received enormous attention, as energy is the limiting resource in these platforms. In the absence of hardware measurements we must resort to models that predict power usage based on events seen by the OS. The accuracy of these models depends on the type of model and on the set of events used as inputs. By looking in detail at the power draw of a wireless card in a laptop running Linux, in this paper we assess the quality of the models one can obtain using events from different levels of the software stack.

In recent work, Pathak et al. advocated the use of system-call tracing, together with an FSM model of components, for predicting power in smartphones. While their technique presents significant advantages over previous counter-based models, we show two concrete examples in which using events at the system-call layer will lose significant accuracy compared to models using events from the device-driver layer. We examine the tradeoffs of using events from the two layers, and argue that device drivers should export a narrow interface that exposes underlying power states to enable simple, accurate, and universal power models by the operating system.

(complete text in pdf)