Reactive Programming Optimizations in Pervasive Computing

Pervasive computing systems are begging for programming models and methodologies specifically suited to the particular cyber-physical nature of these systems. Reactive (rule-based) programming is an attractive model to use due to its built-in safety features and intuitive application development style. Without careful optimization however, reactive programming engines could turn into monstrous power drains of the pervasive system and its sensor network. In this paper we propose two optimizations for reactivity engines. The first, which we prove to be optimal, assumes all sensors in the space are equally important to the application. The other, which is adaptive, employs and estimates a probability for each sensor based on application usage. Both optimizations use a mixed push/pull approach to achieve optimal or near optimal energy efficiency. We present an experimental evaluation of the two algorithms to quantify their performance over a range of parameters.