Timely processing of real-time transactions using fresh temporal data, which reflect the current real world status, is essential in data-intensive real-time embedded applications such as traffic control, agile manufacturing, and defense applications. An overarching goal of this research is developing a novel theoretic framework and prototypical RTEDB (Real-Time Embedded Database) system that can support the desired transaction timeliness and data freshness even in the presence of data/resource contention and dynamic workloads. This project is developing effective Quality of Service (QoS) metrics, novel control theoretic models, feedback control, and feed-forward approaches to specify the desired QoS, model RTEDB dynamics, and support the desired QoS based on the mathematical model, while preventing severe data conflicts. The research is developing several new workload adjustment techniques including adaptive temporal data update, adaptive concurrency control, admission control, and service differentiation techniques to gracefully adapt and differentiate real-time data service with minimal impacts on the user perceived QoS. To improve the scalability of real-time data service, the research focuses on load balancing and temporal data replication in a control area. The research also is developing novel approaches for efficient, secure data transmission in real-time embedded applications. Moreover, the research team is committed to integrating this leading-edge research with teaching and outreach programs. The project, if successful, ultimately envisions that QoS-aware RTEDBs will be used in a number of important data-intensive real-time embedded applications.
