Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Electrical and Computer Engineering


Dr. Miriam A. M. Capretz


The rise of mobile technologies and the Internet of Things, combined with advances in Web technologies, have created a new Big Data world in which the volume and velocity of data generation have achieved an unprecedented scale. As a technology created to process continuous streams of data, Complex Event Processing (CEP) has been often related to Big Data and used as a tool to obtain real-time insights. However, despite this recent surge of interest, the CEP market is still dominated by solutions that are costly and inflexible or too low-level and hard to operate.

To address these problems, this research proposes the creation of a CEP system that can be offered as a service and used over the Internet. Such a CEP as a Service (CEPaaS) system would give its users CEP functionalities associated with the advantages of the services model, such as no up-front investment and low maintenance cost. Nevertheless, creating such a service involves challenges that are not addressed by current CEP systems. This research proposes solutions for three open problems that exist in this context.

First, to address the problem of understanding and reusing existing CEP management procedures, this research introduces the Attributed Graph Rewriting for Complex Event Processing Management (AGeCEP) formalism as a technology- and language-agnostic representation of queries and their reconfigurations. Second, to address the problem of evaluating CEP query management and processing strategies, this research introduces CEPSim, a simulator of cloud-based CEP systems. Finally, this research also introduces a CEPaaS system based on a multi-cloud architecture, container management systems, and an AGeCEP-based multi-tenant design.

To demonstrate its feasibility, AGeCEP was used to design an autonomic manager and a selected set of self-management policies. Moreover, CEPSim was thoroughly evaluated by experiments that showed it can simulate existing systems with accuracy and low execution overhead. Finally, additional experiments validated the CEPaaS system and demonstrated it achieves the goal of offering CEP functionalities as a scalable and fault-tolerant service. In tandem, these results confirm this research significantly advances the CEP state of the art and provides novel tools and methodologies that can be applied to CEP research.