Distributed rule monitoring in distributed active database

Recently, active databases have emerged from the integration of databases and production systems by embedding production-style rules in database management systems. Production systems need the capabilities of a database management system to manage the ever-increasing amount of data and rules to support more and larger applications. Database management systems, in contrast, gain a unifying mechanism for numerous database functionalities which are used to be implemented in an ad hoc manner. When production-style rules are supported, a database actively takes predefined actions (i.e., without user intervention) in response to situations that are specified on the state of the database. Traditional databases, on contrast, react passively to queries or transactions explicitly invoked by users or applications.;As distributed databases gain popularity, the features provided in active databases by production-style rules become desirable in distributed databases. New problems arise with the distributed environment, so do the opportunities for parallelization.;In order to successfully integrated production-style rules into a distributed database, three design issues have to be addressed: Decomposing rules, Distributing rules to sites, and Evaluating distributed rules correctly. The first two issues have been addressed previously in different contexts, but the third issue has just been identified. This dissertation addresses these three issues together using an integrated approach.;Proposed is a rule decomposition algorithm utilizing algebraic manipulations of a relational operator tree as derived from the definition of a given rule. A new relational operator, AND, is proposed to identify independent parts of a rule query, which facilitates the rule distribution later and cuts down the amount of data transferred among sites.;Due to the problems introduced by geographical separation of sites in a distributed database system, such as communication delay and lack of a global clock, correct evaluation of a distributed rule is difficult. Two distributed rule evaluation algorithms, which collect consistent local results from sites to form the global view, are proposed. The performance of the distributed rule evaluation algorithms are studied in terms of message count and response time using analytical models and simulations. The performance data provide feedback for the first two algorithms. Comparisons between the first approach and the rule monitoring algorithms in centralized active databases and between the two proposed approaches are also investigated.