In real-time database systems, transaction processing must satisfy not only the database consistency constraints but also timing constraints such as deadlines. For the most part, database researchers have not dealt with timely issues while the real-time research community has ignored the problem of guaranteeing the consistency of shared data. This work involves the development of an integrated approach for supporting real-time transactions. The following topics will be investigated in depth: (1) Concurrency control protocols, (2) Recovery techniques, (3) Deadlock resolution strategies, and (4) CPU scheduling algorithms. In addition, we will investigate two secondary issues: I/O scheduling and time constrained communication. While we plan to employ simulations and analytical methods where appropriate, to a large extent, all our protocols and algorithms are to be implemented, evaluated and compared on an existing distributed database testbed called CARAT. This is significant because CARAT has the ability to account for the (usually neglected, but not negligible) overheads involved in supporting distributed transactions. This research will have an impact on both the topics with which it overlaps, namely database systems and real-time systems. Also, because of the ubiquitous need for the timely processing of information in all applications that employ computers, such as in banking, command and control, automated factories, and process control systems, this project is expected to produce results that will positively affect many application areas as well.
