Structural complexity in service-oriented systems and its effects on system evolution and performance : a multi-cases study

Abstract

In a dynamic, uncertain environment where flexibility and responsiveness are key, companies are trying to adapt. In their quest for agility, they adopt new processes but also new architecture principles. Among them, Service-Oriented Architecture was meant to allow enterprise systems to become modular. Enhanced modularity would in turn lead to better agility. Yet the correlation between SOA adoption and improved agility performance remains to be seen. Breaking down systems into services is indeed not enough to achieve modularity. This research explores the effects of the structural complexity of composition of services on their performances and capacity to evolve. It does it by conducting a multi-case study on a set of web services responsible for complex tasks in a large and complex system. Drawing from the existing literature, structural complexity is broken down in two dimensions: components and interactions complexities. The effects of both dimensions are studied. While this thesis could not prove a direct effect of structural complexity on robustness and time to deliver, it shows strong evidence of an adverse effect of structural complexity on response time, coordination needs and ability to understand the system. This thesis also looks at the design decisions that can lead to complexity, proposes an enhancement of the structural complexity computation by looking at shared data and suggests conceptual integrity as a necessary complement to a pure structural analysis. The paper concludes with suggestions for potential future research.