Combining Expressiveness and Efficiency in a Complex Event Processing Middleware

Several complex systems operate by observing a set of primitive events that happen in the external environments, interpreting and combining them to identify higher level composite events, and finally sending the notifications about these events to the components in charge of reacting to them, thus determining the overall system’s behavior. Examples of systems that operate this way are sensor networks for environmental monitoring, financial applications, fraud detection tools, and RFID-based inventory management. More in general, as observed in [1], the information system of every complex company can and should be organized around an event-based core that realizes a sort of nervous system to guide and control the operation of the other sub-systems. The task of identifying composite events from primitive ones is performed by the Complex Event Processing (CEP) Engine. It operates by interpreting a set of event definition rules that describe how composite events are defined from primitive ones. The CEP engine is usually part of a CEP system or middleware which also handles the communication with local and remote clients. To capture all the requirements of the aforementioned applications, a CEP engine has to face several challenges. First, it has to provide a suitable language for rule specification, explicitly tailored to model complex temporal relationships that join together primitive events in composite ones. Second, it has to implement efficient processing algorithms, to detect composite events and deliver notifications to interested parties with the lowest possible delay. Finally, it has to support distributed scenarios, in which the communication parties may be deployed over a wide geographical area. This thesis first proposes a modelling framework to compare and analyze not only existing CEP systems, but all the systems developed with the aim of processing continuous flows of information according to predeployed processing rules. This allows us to identify the main advantages and limitations of existing approaches, by looking at a wide range of proposals. Moreover, our modelling framework draws a common ground for comparing efforts coming from different research communities, with different background, expertise, and vocabulary. We believe that our work can bridge the gap between different worlds, promoting the communica-