The Microstructure of Complex Architectures in Open Collaboration: A Theory of Design Network Motifs
26 Pages Posted: 29 Apr 2020
Date Written: December 17, 2019
The established stream of literature on design architectures argues that designers should aim for modular architecture in order increase the system’s technical performance by lowering the propagation costs of the system’s design: A system with a small stable “core”, a cycle of coupled parts of the system, and a large variable periphery reduce the risk of technical feature and increase opportunity for innovation. However, such a core-periphery view ignores the micro-level dependency structures that emerge in open collaboration when a large number of developers produce a complex technical system at distance, virtually, and outside of formal employment relationships. In this paper, we develop a theory of design network motifs that accounts for the smallest design dependencies in the system that emerge in situ. Informed by the network theoretical concept of a “motif”, we introduce design motifs as the smallest unit of design interdependencies in a system architecture and develop a theory of design network motifs describe each system’s ‘motif fingerprint’ and explains how certain motif patterns affect the performance of a system as a whole. We empirically examine our design network motif theory using architectural data of the Open Stack repository NOVA, consisting of 2359 files and 1545532 lines of code. We extract all potential design network motifs created by 872 developers for 5 years. Our results show that design network motifs offer a new way to explain hidden dependencies in complex system architectures. Surprisingly, it is not just cycling that significantly impact a system’s performance but instead, a “feed-forward loop” motif. Our results contribute to the literature on design architectures and open collaboration more broadly.
Keywords: architecture, modular system theory, loose coupling, propagation cost, DSM, network motifs, open collaboration
Suggested Citation: Suggested Citation