2005-06-30Buch DOI: 10.18452/2445
Parallel event detection in active database systems
the heart of the matter
This paper proposes a strategy for parallel composite event detection in Active database systems (DBS). Up to now, the detection is sequential and totally synchronized, and thus preserves the timely order of events during the detection process. However, in distributed and extensible applications events may occur simultaneously in parallel unsynchronized streams. In order to adapt composite event detection to those new requirements we relax the timely order of events to a partial order and process parallel event streams. As a consequence, composite event detection must deal with unsynchronized and parallel event compositions. This paper investigates several parallel event detection strategies. We adapt techniques and optimizations from query execution in Relational DBS. However, queries and composite events differ fundamentally. The semantics of query execution is set oriented, whereas composite event detection bases on open streams of events. Cardinality and frequency of incoming events is unpredictable; timely order is essential. Parallel execution destroys the timely order of events and compositions. Our approach introduces a hybrid parallelization strategy for composite event detection in Active DBS that respects the timely order of events.
Files in this item