Paper detail

Low Power Dynamic Scheduling for Computing Systems

This paper considers energy-aware control for a computing system with two states: "active" and "idle." In the active state, the controller chooses to perform a single task using one of multiple task processing modes. The controller then saves energy by choosing an amount of time for the system to be idle. These decisions affect processing time, energy expenditure, and an abstract attribute vector that can be used to model other criteria of interest (such as processing quality or distortion). The goal is to optimize time average system performance. Applications of this model include a smart phone that makes energy-efficient computation and transmission decisions, a computer that processes tasks subject to rate, quality, and power constraints, and a smart grid energy manager that allocates resources in reaction to a time varying energy price. The solution methodology of this paper uses the theory of optimization for renewal systems developed in our previous work. This paper is written in tutorial form and develops the main concepts of the theory using several detailed examples. It also highlights the relationship between online dynamic optimization and linear fractional programming. Finally, it provides exercises to help the reader learn the main concepts and apply them to their own optimizations. This paper is an arxiv technical report, and is a preliminary version of material that will appear as a book chapter in an upcoming book on green communications and networking.