Concepts and tools for measuring the complexity of service choreography models

Abstract

With the increasing acceptance of Service-oriented Architectural (SOA) style to deliver services over the network by businesses in todayís competitive world have shaped the two major paradigms to realize service compositions: choreographies and orchestrations. Services are almost ubiquitous in carrying a unit of software functionality to the end user. Although both the paradigms of service composition, namely orchestration and choreography, are devised to execute a conversation protocol between the concerned participants, nevertheless the service choreography designed using a modelling language is quite convoluted. With the increase in importance of data in modern computing, TraDE concepts were adapted for data-aware service choreographies to decouple the data flow from the intricacies of the control flow associated in the message exchange between participants of the choreography. The TraDE middleware assists in controlling the data transfer among participants in a straightforward and automated manner. The service choreography models in this thesis are framed using BPMN 2.0 and BPEL4Chor modelling languages for studying their degree of complexity. In this research work we analyse existing complexity metrics (custom built for process models) and to which extent they are efficient in predicting the complexity of service choreographies that are modelled by incorporating the TraDE approach. We devise our complexity metrics based on adaption from software engineering, graph theory, and business process model research domain to measure the level of complexity for TraDE concepts applied service choreography model. The complexity metrics we adapted for applying to the service choreography model are Size Metrics, McCabeís Cyclomatic Complexity, Control-flow Complexity, Interface Complexity, Coefficient of Network Complexity, Durfee Square and Perfect Square Metric, Connectivity Level between Activities, Halstead-based Choreography Complexity, and Structural Metrics. We also develop the prototypical implementation of the framework for computing all the devised complexity metrics applied onto a user specified service choreography model and persistently store the generated metric results. The developed framework is evaluated for its efficacy in computing the complexity metrics on a collection of service choreography models that are designed in both BPMN 2.0 and BPEL4Chor modelling notations.