Energy-efficiency is vital for pervasive computing. Small, mobile computers such as handhelds are battery powered. They must carefully manage their limited energy supply in order to run demanding applications for long periods of time. Pervasive computers are also increasingly well-connected to the Internet by wireless networks. However, connectivity is costly: communication is a large and growing portion of their power budgets. Current data access mechanisms exacerbate the problem---they substantially increase the energy cost of communication because they are optimized for performance, not energy-efficiency.
This research investigates how to build energy-efficient distributed file systems for pervasive computing. The first step is building measurement infrastructure that precisely captures the communication, storage, and computation costs of data access. The next step is characterizing the energy usage of existing distributed file systems and identifying characteristics that increase energy-efficiency. The final step is building a new, energy-efficient file system that enables pervasive computers to maintain connectivity without compromising battery lifetime. This research thus allows pervasive computing devices to be incorporated into everyday activities. For example, all-day battery life is essential for projects that deeply integrate handheld computers into classroom activity.
