State-Filters for Enhanced Filtering in Sensor-Based Publish/Subscribe Systems

Publish/subscribe systems have been extensively studied in the context of distributed information-based systems, and have proven scalable in information-dissemination for many distributed applications that have motivated the research. With the emergence of sensor-based applications and sensor networks, researchers have proposed novel publish/subscribe protocols that address the problem of distributed event dissemination for sensor network characteristics and constraints. In this paper, we focus on primitive events and the emerging class of publishers, and argue for "state-filters" as more useful and suitable means of filtering events (than content-based filtering) in sensor-based publish/subscribe systems. Using state-filters, we claim to achieve higher efficiency by means of filtering redundant and correlated event notifications, suppress event duplicates, and capture lasting conditions that had been previously not possible using content- based filters. We evaluate our proposed filtering mechanism using real-world sensor data, and highlight some assumptions and pitfalls that motivate our future work in this area.