Action-Oriented Query Processing for Pervasive Computing

Wenwei Xue

HKUST

Pervasive computing applications monitor physical-world phenomena and integrate data acquisition, communication and operations across small, heterogeneous devices (e.g., smart sensors, network cameras and handheld devices). To ease the development of these applications, we propose to perform their tasks by executing action-embedded queries, which are event-driven continuous queries with operations on devices (actions). We extend SQL to allow applications to specify actions and action-embedded queries. We develop a uniform data communication layer that enables network data independence over a number of heterogeneous devices. We treat actions as first-class citizens (operators) in query execution plans, and investigate adaptive, cost-based optimization techniques for a single query as well as for multiple queries. We evaluate the performance of our prototype query processor, Aorta, using a pervasive lab monitoring application. Results from empirical studies as well as simulation studies show that Aorta ensures correct application semantics, improves query response time and balances device workload.