Handling sensor data in rapidly changing environments to support soft real-time requirements

Databases have been mandated as the chosen platform to develop automated control systems for environments where the state of the system changes rapidly and actions need to be taken with timing constraints, for example, stock trading, air traffic control, network management, and process control to name a few. Input data in such systems come from sensors distributed throughout the physical environment being monitored. As the system state changes rapidly, the sensors report information fast and with high volumes. Control decisions in such systems are highly dependent on the accuracy and timeliness of such recorded sensor-reported data. In this paper we suggest ways to algorithmically maintain the accuracy and timeliness of sensor data by enforcing two porperties termed recency and temporal consistency. Simultaneously, our algorithms also attempt to satisfy quality of service requirements on response times of queries submitted in such systems. The performance of these algorithms is validated through an extensive simulation.