A rule-based cooperative transaction model and event processing in real-time active database systems

This thesis proposes a model for database transactions to cooperate through rules, task communication and synchronization. Database transactions manipulate persistent and transient data objects. With event-condition-action (ECA) rules from active databases, the model provides transactions with a new way of cooperation. Task communication and synchronization can be specified not only statically inside transactions, but also dynamically through rules. Statically specified task communication and synchronization cannot handle dynamic changes, which provides the motivation for using rules. A transaction manager is involved in transaction cooperation by invoking transaction manipulation operations, e.g., start a transaction as a result of an event.;Three new types of non-ECA rules (that cannot be expressed using ECA rules) are introduced. For example, a non-ECA rule can prohibit an action from being performed in a certain time interval on the occurrence of an event. Transaction cooperation in real-time database systems with time constraints are considered using both ECA and non-ECA rules. Events and actions (transactions) of rules have temporal constraints. Rules specify the time intervals to perform or to prohibit actions upon the occurrence or the absence of events, possibly in given time intervals. Algorithms for enforcing non-ECA rules with time constraints are provided in the thesis. Given a set of rules and an event occurrence, performing and prohibiting an action at the same time may be deduced, which clearly implies inconsistent rules. The consistency property of a set of rules is investigated.;The transaction model is used in real-time active databases with repeating events, and event-action rules with priorities. In order to avoid missing action (transaction) deadlines, a mechanism is needed to select rules to fire (in the presence of currently executing actions). Such a mechanism executes at fixed time intervals, called the Rule Selection Period. Three desirable properties are identified: (a) Completion Property: triggered actions of fired rules (and currently executing actions) complete within their deadlines with a probability higher than a given threshold, (b) Duration Property: all triggered actions are expected to complete within the Rule Selection Period, and (c) Priority Property: sum of the priorities of fired rules are maximized.;The thesis develops a probabilistic model, and an inexpensive (polynomial-time) algorithm for the satisfaction of the Completion Property. Satisfying the Priority and the Duration Properties at the same time is shown to be NP-Complete. Therefore, two heuristics are given, and their performances are compared and evaluated extensively by simulation.