Complex Event Processing deals with aggregating simple events, which are defined as "occurrences of significance in a system", to get semantically-richer events usable by an end application. They have been studied earlier in multiple disparate contexts, for example, in the Active Database community, under ECA rules, that are used to build triggers for a variety of purposes.

A resurgence of interest in Complex Event Processing research has taken place because of recent advances in sensing technologies like sensornets and RFID. This technology generates events out of real-world inputs such as the movement of people in a home or items in a supply chain. Our work focuses on Complex Event Processing in the context of such real-world event sources, and on two challenging dimensions that arise: streaming data and fuzzy or probabilistic data. We present the notion of "semantic windows", which go beyond time-based or tuple-based windows proposed for streaming data processing. Probabilistic Complex Event Processing (PCEP) allows applications to reason about and respond to events in scenarios where simple events cannot be monitored in a crisp fashion.

Ideas from this work have been implemented in the TelegraphCQ streaming data processor, and used to drive the core functionality of an event-driven library scenario in a recent system demonstration.




Download Full History