Context-aware Event Stream Analytics

Complex event processing is a popular technology for continuously monitoring high-volume event streams from health care to traffic management to detect complex compositions of events. These event compositions signify critical “application contexts” from hygiene violations to traffic accidents. Certain event queries are only appropriate in particular contexts. Yet state-of-the-art streaming engines tend to execute all event queries continuously regardless of the current application context. This wastes tremendous processing resources and thus leads to delayed reactions to critical situations. We have developed the first context-aware event processing solution, called CAESAR, which features the following key innovations. (1) The CAESAR model supports application contexts as first class citizens and associates appropriate event queries with them. (2) The CAESAR optimizer employs context-aware optimization strategies including context window push-down strategy and query workload sharing among overlapping contexts. (3) The CAESAR infrastructure allows for lightweight event query suspension and activation driven by context windows. Our experimental study utilizing both the Linear Road stream benchmark as well as real-world data sets demonstrates that the context aware event stream analytics consistently outperforms the state-of-the-art strategies by factor of 8 on average.