Doctor of Philosophy
School of Computer Science and Software Engineering
Ponnalagu, Karthikeyan, Goal oriented variability management in model driven service engineering, Doctor of Philosophy thesis, School of Computer Science and Software Engineering, University of Wollongong, 2016. http://ro.uow.edu.au/theses/4964
Service Oriented Applications (SOA) are designed and developed as a collection of interdependent models using the approach of Model Driven Architecture (MDA). Subsequent reuse of such models in different states of abstraction requires them to be modeled with orthogonal dimensions of variability. The key objective of adopting service orientation in application development is to maximize reuse of both strategic and operational type business functions. Such business functions are developed in the form of processes and services across multiple applications required for different usage contexts. This thesis presents a goal oriented methodological framework to address variability management in service oriented systems.
Variability Management (VM) as a research field has received considerable academic as well as industry attention in the last few years. Most of the current approaches, associate variability tightly in the context of new system design and development. For example, the techniques for validating the readiness of existing software systems towards supporting variability has been mostly ignored. This can result in erroneous and redundant modifications. Therefore, there is significant research space that left to be convincingly covered to develop formal techniques for context-driven variant generation, goal-driven variant validation, identifying candidate variants from asset repositories and so on.
Goal decomposition models are always found to be effective in capturing and relating requirements stated by different stakeholders of a business organization. A goal represents of intended operational states that an organization seeks to realize in a formalized approach. In this research work, goal is treated as a binding guidance for designing or adapting multiple variants associated with a process design. The aspect of goal preserving variations has not been addressed adequately in existing research on variability management.
As a goal model efficiently captures multiple stakeholder expectations on a given system, it must be ensured that both the discovered and derived variants of a given system are goal preserving in nature. This requires the ability to formally reason the correlation and differentiation of such variants in terms of their semantic alignment with a given set of goal models.
This thesis presents a novel framework that supports variability management in SOA-based systems. In particular, a formal mechanism called Goal Oriented Variability Management (GOVM) for reasoning, representing and realizing possible goal preserving variations for a given system is provided. It successfully demonstrates how semantically annotated process models in BPMN notation and goal models in UML notation can be leveraged in an integrated fashion. It also provides a lattice based algebraic approach to formally organize a collection of variants with or without partially ordered relationships (Containment, Composition, contextual, Dependency, goal alignment) between them. This thesis further demonstrates formal techniques for goal driven process derivation, effective reuse of business processes and a formal representation of service variability model. Experiments with an extensively developed implementation as an eclipse based application framework demonstrate the feasibility of the core proposed techniques of this work.