Towards SOA implementation complexity measurement enlightened by organization theory.
Abstract
When implementing information infrastructures to support business, Service-Oriented Architecture (SOA) based systems are inevitably more complex than traditional architecture based systems. Therefore, it is essential to measure the implementation complexity to avoid development that is unworthy of adopting SOA. However, there is little work that covers the SOA implementation complexity measurement with comprehensive analysis. By adopting existing work related to organizational complexity, this paper proposes a framework for comprehending and measuring the organizational complexity of SOA implementation, which comprises four dimensions: Structure, Environment, Business and Resource. Along each dimension, some SOA-compatible metrics have been extracted from the corresponding research on organizational complexity. More importantly, this proposed framework can be used for future work in proposing metrics for integrally and quantitatively calculating the complexity of implementing SOA. Categories and Subject Descriptors D.2.8 [Software Engineering]: Metrics - complexity measures. D.2.9 [Software Engineering]: Management - cost estimation. F.1.3 [Computation by Abstract Devices]: Complexity Measures and Classes - relations among complexity classes. H.5.3 [Information Interfaces and Presentation]: Group and Organization Interfaces - organizational design. General Terms Measurement, Design, Theory Keywords Service-Oriented Architecture (SOA), Organization Theory, Organizational Complexity, SOA Implementation Complexity, Measurement. 1. INTRODUCTION Service-Oriented Architecture (SOA) has become a widely accepted and adopted architectural style that provides the benefits of agility, flexibility and adaptability for dynamic businesses. However, the loosely coupled asynchronous SOA systems are inherently more complex than the traditional architecture based systems. Josuttis [16] has pointed out that distributed processing would be inevitably more complicated than non-distributed processing, and any form of loose coupling increases complexity. In practice, building a true heterogeneous SOA for a wide range of operating environments may take years of development time if