Managing Rules In Active Databases PDF Download

Abstract: "Traditional databases are viewed as passive repositories of data. We store, retrieve, and alter data in a database, but do not expect a database to respond to actions performed to it. This passive paradigm is now rapidly changing as databases incorporate rules which permit them to detect and respond to important events. The resulting active databases have expanded capabilities. Among other things, they can enforce constraints, maintain derived data, propagate the effect of updates, and implement change management protocols. The incorporation of rules, and their active functionality, must consider existing database functionalities such as safe concurrent access to data, secure access to data, and efficient access to large amounts of data. Each functionality must be integrated with the rule model. In this thesis, we consider the effect of incorporating active database rules on the latter two functionalities: secure and efficient access to data. Specifically, we present MLS rules, an integration of the event-condition-action (ECA) rule model and the multilevel secure (MLS) relational data model. MLS rules are secure, according to MLS constraints, yet retain most of the powerful functionality of rules. Covert timing and storage channels are considered in the design of MLS rules. For example, cascading rule processing is used to help hide secure rule executions. In order to perform this integration, it was first necessary to develop a semantically-based MLS relational model, as existing models were found semantically indadequate to support the integration of rules. This work is presented in this thesis. We define entity-relationship modeling in the MLS model, multilevel relations, and semantically-based query and update operations. Finally, we address the problem posed by runaway execution of rule chaining. We address the optimization problem of eliminating redundant rule execution during chaining, introducing a methodology based on a combination of static analysis of the rule set, and the dunamic management of executing rules. This join number methodology eliminates redundant execution while incurring a very low overhead for rule management."