| This research addresses real-time data management in the distributed environment. It includes the design of real-time databases which must function in the distributed environment and related real-time task scheduling problems. It has been generally acknowledged that traditional database design is inadequate for handling time-critical transactions. In particular, the usual ACID semantics of traditional databases is too restrictive for many real-time applications. On the other hand, the classical task models considered by the real-time computing community are inadequate for capturing the characteristics of distributed real-time transactions because of highly variable timing parameters.; The goal of this research is to build a real-time database engine on top of a set of heterogeneous databases in the distributed environment. There are two main components in this work. We concretize the concept of “similarity” to “real-time similarity predicates” which are incorporated into database queries. By doing so, we can adjust the computational requirements of queries to trade off between computation and communication load. This is of the utmost importance in large networks such as those in geographically dispersed simulation systems. We also investigate a suite of task models that are extensions of the classical model of Liu and Layland. The scheduling of distributed real-time tasks and multimedia transmission are investigated. Our extension will permit flexible timing parameters in real-time tasks such as queries that incorporate similarity constraints. The practicality of our ideas will be demonstrated by the design of a prototype real-time database engine. |
