Intellectual Merit: The power consumption rate of computing devices has been increasing exponentially. This makes it increasingly difficult to supply energy to devices and to cool these devices. Poweraware computation is especially important in the domain of sensor networks which are composed of small battery-powered nodes. Power management in sensor networks is viewed as so critical that it must be dealt with at all layers of the protocol stack.
Many power management techniques have been proposed and implemented. Most of these techniques are similar in that they reduce or eliminate power to some or all components of the device. However, there is an inherent conflict between power management and performance; in general, the more power that is available, the better the performance that can be achieved. As a result, it is generally proposed that power reduction techniques be preferentially applied during times when performance is less critical. However, this requires a policy to determine how essential performance is at any given time and how to apply a particular power reduction technique. For example, to use the frequency scaling technique, where the speed of the clock is changed dynamically, one needs a policy to set the speed at each point in time. There is a growing consensus that these policies must incorporate information provided by applications and high levels of the operating system, and that current tools and mechanisms for power management are inadequate and require more research. The authors propose to formalize power management problems as optimization problems, and then develop algorithms that are optimal by these criteria. The goal of this research is to develop effective algorithms for specific problems within the domain of power management, as well as to build a toolkit of widely applicable algorithmic methods for problems that arise in energy-bounded and temperature-bounded computation. The authors propose to initially focus on problems that deal with speed scaling and power-down techniques, since these are currently the dominant techniques in practice.
Broader Impacts: The authors propose to both develop fundamental theoretical techniques, and to apply these techniques to attack timely and important applications in computer systems. Both PIs have an established track record of working closely with researchers in applied areas to ensure that the theoretical models developed match the associated real-world problems. This is essential for theoretical results to have an impact. This work will continue to foster this very productive cross-fertilization between these experimental systems researchers and theoretical computer science. The students supported under this grant will be influenced by this philosophy of research. They will be trained to be proactive in working with researchers in applied domains to bring important and interesting problems into the theory community. They will also be encouraged to publish the resulting work in systems as well as theory conferences to ensure that new algorithmic discoveries have an impact. As part of this project, the authors also plan to continue outreach work to high schools students, encouraging underrepresented groups to choose careers in technology related fields. They have developed a talk outlining diverse opportunities within computer science and plan to involve graduate and undergraduate students in presenting this talk at local high schools. The work in this proposal is featured in the talk. In addition, they plan to involve students from underrepresented groups in research projects related to power management.
