EPA Compositions on Realtime Streaming Analytics

Abstract

Context: Complex Event Processing (CEP) architectures present high applicability in Realtime Streaming Analytics (RTSA) scenarios by the extraction and generation of valuable information from continuous data feeds, like in stock markets, traffic, and patient monitoring. Problem: Although guidelines and models for CEP architectures have been proposed in academia and industry, the composition of its inter-operable elements in charge of processing events, known as Event Processing Agent (EPA), remains a recurring challenge for software architects. Solution: This work proposes EPAComp, a model that covers this gap and addresses large-scale processing requirements through features such as stream-based constructions and specialized EPAs (e.g., for pattern detection). IS Theory: We employed Representation theory in this work towards creating a model that represents information system for event processing. Method: The model was applied in a real case experiment to create a solution to collect user requests, as streams of events, from around 200 systems, and to provide a dashboard for monitoring their usage. Besides, industry experts qualitatively evaluated the proposal. Results: The experiment results show an application of the model to handle heterogeneous data in a scalable and efficient manner according to indicators regarding performance, assertiveness of processed output, and degree of cohesion, and coupling of components. In the qualitative results presents that experts asserted EPAComp capabilities fits RTSA requirements. Contributions: The main contributions is an architectural model for EPA composition that enhances the literature by representing static and dynamic EPA compositions through arrangements of specific aggregation structures; representing state of the art event processing strategies in CEP; and, organize hierarchy of EPA types.