The Complex Event Processing Paradigm

As mentioned in the previous chapters, pervasive systems demand for continuous processing of information collected from multiple sources deployed in the environment under analysis. Often, a large portion of such information encodes notifications of events occurred within the pervasive system or in the environment where it operates. In such cases, the goal of the processing step is to detect situations of interest as soon as they occur, by looking at the primitive events that have been observed. On the one hand, this requires the ability of defining the situations of interest (often called composite events) as patterns of primitive events, joined by specific relationships. On the other hand, this also demands the capability of effectively exploiting such definitions to detect composite events at run-time, as soon as they occur. Complex Event Processing (CEP) languages and systems have been proposed by research and industry to satisfy these complementary needs. The CEP paradigm shares some similarities with the Data Streaming approach described in Chapter 4. Nevertheless, several key differences exist. Specifically, Data Stream Management Systems (DSMSs) focus on transforming the incoming streams of information into new streams, e.g., by joining or aggregating data items. Conversely, CEP systems interpret incoming information as notifications of events and focus on detecting relevant patterns among such streaming notifications. This chapter focuses on this specific processing abstraction, describing the key features and issues that characterize it.