Distributed computing applications, such as groupware and manufacturing automation, have concurrently executing processes that access common shared data. The applications require maintaining the consistency of shared data and correctness of executions, and also, have demanding real-time performance requirements. This research examines, designs, and implements, a system based on time-constrained transaction-oriented distributed computations. Consistent executions are maintained using modified database concurrency control, and improvement of performance is addressed using techniques from real-time data management. Building on existing research results, the system incorporates several correctness and efficiency criteria, and is flexible to changes and scale-up. Learning from on-going large-scale efforts in the application areas, the approach is tested by implementation of substantial functionalities for the target domains. Real-time scheduling algorithms are incorporated into a main-memory data management system, and the primitives for distributed concurrency control using synchronization protocols are implemented across multiple instances of such systems. This approach is to be empirically compared in terms of functionality and performance against the systems already available. The complementary education plan includes fostering off-line group study and learning by doing hands-on work. Students, organized into self-study groups, examine facets of the research problems, and obtain broad guidelines from information technology tools. Also, the research problems are assigned as project work which fosters a hands-on approach to learning. In the long term, this research and educational dissemination will influence the design and development of large-scale real-time distributed applications using database and transaction technologies.
