Résumé: Most work in sensor networks tries to maximize network lifetime. However, for many applications the required lifetime is known in advance. Therefore, application quality should rather be maximized for that given time. Levels, the approach presented in this article, is a programming abstraction for energy-aware sensor network applications that helps to meet such a user-defined lifetime goal by deactivating optional functionality. With this programming abstraction, the application developer defines so-called energy levels. Functionality in energy levels is deactivated if the required lifetime cannot be met otherwise. The runtime system uses data about the energy consumption of different levels to compute an optimal level assignment that maximizes each node’s quality for the time remaining. As described in this paper, Levels includes a completely distributed coordination algorithm that balances energy level assignments and keeps the application quality of the network roughly constant over time. In this approach, each node computes its schedule based on those of its neighbors. As the evaluation shows, applications using Levels can accurately meet given lifetime goals with only small fluctuations in application quality. In addition, the runtime overhead both for computation and for communication is negligible.

Langue: Anglais