Service-oriented architecture (SOA) has been promoted as a technology that can enhance information systems agility, interoperability between applications, deployment flexibility, and reusability. As with any new information technology (IT), the decision to adopt SOA cannot be taken lightly, given the nontrivial investment in economic and personnel resources. The complexity associated with industry-wide diffusion, coupled with organization, industry, and environment factors, contributes to a lack of a clear strategy for assessing the business value that SOA provides an organization. This research attempts to shed light on this process of value creation for an organization, using a system dynamics approach. A detailed model of the industry diffusion coupled with organization adoption is presented. After suitable calibration and validation, a series of simulations using the model evaluate the efficacy of SOA under a variety of diverse conditions. The results of the simulations indicate clear benefits of SOA over monolithic ITs when employed under appropriate conditions. Situations where SOA fails to live up to expectations are also identified. The model and accompanying simulations can serve as a practical decision support tool for an organization to help make the strategic decisions of adopting and implementing SOA.
With the move to distributed systems and an increasing emphasis on the use of object-orientation for new system design, effective distribution of object-oriented applications is becoming an important concern for designers. Early research in this area has focused on object-clustering schemes for shared memory configurations that have limited value to business applications, which must be distributed over loosely coupled networks. These applications also exhibit the properties of simpler structural relationships and a large number of instances, demanding approaches closer to fragmentation and allocation instead of clustering. This paper develops an approach to distribution of object-oriented applications over geographically dispersed sites in loosely coupled networks---taking account of concerns such as encapsulation, inheritance, messaging, and implicit joins. The approach consists of two phases. First, we develop a scheme for generating class fragments, which ensures that encapsulation is not violated and inheritance is not stretched across sites. Second, considering the message-intensive operation of object-oriented systems, we devise models for allocation of class fragments to sites that minimize inter-site traffic. A nonarbitrary procedure to compile traffic volume estimates exploiting the notion of implicit joins in object-oriented applications provides the natural linkage between the two phases. A research prototype was implemented to establish feasibility of the proposals. We demonstrate usefulness of the approach by its application for distribution of a real-world information system.