| Active database technology enhances traditional databases with rules that are executed in response to database events. This enhancement promises exceptional returns. Useful applications of the technology include view maintenance, workflow, and real-time decision control systems.; Unfortunately, penetration of active databases has largely been restricted to simple rule systems. This narrowed focus can be attributed to an explicit connection of the active rules to the underlying concurrency control system. Although this explicit connection provides a level of flexibility, it becomes semantically intractable in more complex applications. Furthermore, rule execution is computationally intensive. Since rules spawn the execution of many queries over the contents of the database, substantial skill is required to develop complex applications without introducing long duration transactions.; This dissertation addresses three issues surrounding application development that inhibits the feasibility of active databases. First, a quantitative evaluation is performed on the semantics of VenusDB, a modular active database language that executes within the nested transaction model. This evaluation provides evidence that rule modules improve system maintainability. Second, the most general contribution of this dissertation is the identification and study of Log Monitoring Applications (LMAs). LMAs are expert system applications that analyze logs maintained in a database. This dissertation develops the formal execution semantics and correctness proofs of concurrency schemes for applications within the LMA class. The results show that only a minimal number of coupling modes are necessary for the database integration of hard rule systems obeying the LMA restrictions. Third, the architecture and evaluation of an active database optimizer is presented. This optimizer uses database statistics to optimize rules as well as suggest physical schema optimizations. The optimizer is integrated within VenusDB to improve system scalability. |
