Emerging structures in information systems: a SOS approach

Large industrial information systems are composed of many applications, custom-developed or COTS. Each application supports a specific business activity, and business processes are typically implemented by a distributed network of applications that evolves and grows over time. Usually, different development teams or vendors are responsible for the evolution process, each working on only part of the whole system. As a result, the organization suffers from knowledge fragmentation leading to an information system that easily evolves into a chaotic structure. In this paper, we present an experience report of developing and using SoS models of an industrial information system. Based on the insight that information systems are, by all means, systems of systems (SoS), we suggest that this view may mitigate the above stated problems. We model the information system at the SoS level of abstraction. The model appears highly complex and unmanageable. The reason is that the SoS is not designed by an external entity (a system architect) but is the result of a growing process. We research if structures are emerging from the behavior-driven evolution and detect business clusters - sets of applications/systems implementing the processes of a business area. We classify the business clusters according to their shapes and apply the concepts of cohesion/coupling to them, deriving cluster measures. We discuss some possible uses of the clusters and describe how different stakeholders really used them and have benefited from the model. As these structures are at an intermediate level of abstraction between single applications/systems and the whole information system, they help addressing the issue of knowledge fragmentation.