Integration of Event Processing with Service-oriented Architectures and Business Processes.
Technische Universität, Darmstadt
[Ph.D. Thesis], (2014)
dissertationAppel-tuprints.pdf - Accepted Version
Available under Creative Commons Attribution Non-commercial No Derivatives, 2.5.
Download (7MB) | Preview
|Item Type:||Ph.D. Thesis|
|Title:||Integration of Event Processing with Service-oriented Architectures and Business Processes|
Data sources like the Internet of Things or Cyber-physical Systems provide enormous amounts of real-time information in form of streams of events. The use of such event streams enables reactive software components as building blocks in a new generation of systems. Businesses, for example, can benefit from the integration of event streams; new services can be provided to customers, or existing business processes can be improved. The development of reactive systems and the integration with existing application landscapes, however, is challenging. While traditional system components follow a pull-based request/reply interaction style, event-based systems follow a push-based interaction scheme; events arrive continuously and application logic is triggered implicitly.
To benefit from push-based and pull-based interactions together, an intuitive software abstraction is necessary to integrate push-based application logic with existing systems. In this work we introduce such an abstraction: we present Event Stream Processing Units (SPUs) - a container model for the encapsulation of event-processing application logic at the technical layer as well as at the business process layer. At the technical layer SPUs provide a service-like abstraction and simplify the development of scalable reactive applications. At the business process layer SPUs make event processing explicitly representable. SPUs have a managed lifecycle and are instantiated implicitly - upon arrival of appropriate events - or explicitly upon request. At the business process layer SPUs encapsulate application logic for event stream processing and enable a seamless transition between process models, executable process representations, and components at the IT layer.
Throughout this work, we focus on different aspects of the SPU container model: we first introduce the SPU container model and its execution semantics. Since SPUs rely on a publish/subscribe system for event dissemination, we discuss quality of service requirements in the context of event processing. SPUs rely on input in form of events; in event-based systems, however, event production is logically decoupled, i.e., event producers are not aware of the event consumers. This influences the system development process and requires an appropriate methodology. Fur this purpose we present a requirements engineering approach that takes the specifics of event-based applications into account.
The integration of events with business processes leads to new business opportunities. SPUs can encapsulate event processing at the abstraction level of business functions and enable a seamless integration with business processes. For this integration, we introduce extensions to the business process modeling notations BPMN and EPCs to model SPUs. We also present a model-to-execute workflow for SPU-containing process models and implementation with business process modeling software.
The SPU container model itself is language-agnostic; thus, we present Eventlets as SPU implementation based on Java Enterprise technology. Eventlets are executed inside a distributed middleware and follow a lifecycle. They reduce the development effort of scalable event processing applications as we show in our evaluation. Since the SPU container model introduces an additional layer of abstraction we analyze the overhead in terms of performance and show that Eventlets can compete with traditional event processing approaches in terms of performance.
SPUs can be used to process sensitive data, e.g., in health care environments. Thus, privacy protection is an important requirement for certain use cases and we sketch the application of a privacy-preserving event dissemination scheme to protect event consumers and producers from curious brokers. We also quantify the resulting overhead introduced by a privacy-preserving brokering scheme in an evaluation.
|Place of Publication:||Darmstadt|
|Uncontrolled Keywords:||Event Processing, Business Process Management, Publish/Subscribe, Reactive Middleware|
|Classification DDC:||000 Allgemeines, Informatik, Informationswissenschaft > 004 Informatik|
|Divisions:||20 Department of Computer Science
20 Department of Computer Science > Databases and Distributed Systems
|Date Deposited:||10 Jun 2014 07:04|
|Last Modified:||10 Jun 2014 07:04|
|Referees:||Buchmann, Prof. Alejandro and Karastoyanova, Jun.-Prof. Dimka and Eugster, Prof. Patrick|
|Refereed:||20 May 2014|