An Active Component for a Parallel Database Kernel

The Monet parallel database server is an experimentation platform for a variety of datamodels and novel applications. In this paper we describe its active behavior based on the notion of trigger abstractions and an event notification scheme. Trigger abstractions can be used to construct intricate trigger instance networks. The events notified in the DBMS kernel threads flow into a shared event pool. The trigger-event-monitor watches this pool for event combinations to enable trigger firing. We illustrate how these concepts can be used by a rule compiler and describe a performance metric to guide the search for an efficient architectural solution.