My research considers the database facilities that are needed by the new class of applications requiring soft real-time constraints, focusing on two problem in particular. The first problem is how to integrate soft real-time databases (SRTDBs) into distributed systems composed of real-time and non-real-time databases. Supporting data sharing, and distributed materialized views in particular, across such a disparate set of databases presents interesting problems in both scheduling and efficient implementation. Even if these issues are resolved, the data materialized at the SRTDB may not be usable in its raw form: Quite often, users are more interested in aggregate data. Thus the SRTDB must provide support for efficiently computing derived data. This is the second major problem addressed in my thesis.
My dissertation presents solutions to these problems in two phases. First, many of the general tradeoffs of the two problems are studied through simulation, allowing a wide coverage of algorithms and parameter values. Next, a few of the most promising solutions are implemented and compared in an actual database in order to study second order effects and validate the simulation studies. The result is an active real-time main memory database called STRIP (STanford Real-time Information Processor) which I (with Ben Kao) have designed, coded and tested. STRIP provides unique support for importing and exporting data subject to real-time constraints as well as a powerful rule engine that allows derived data to be maintained very efficiently.
Ambrose Bierce's "The Devil's Dictionary"