Related items at Kannisto.org

Research project(s): CODES

Keywords

Industrial SOA; Executable business processes; Life cycle simulation; XML; BPMN; WS-BPEL; Web Services; HTTP; DPWS; Real-time system

Abstract

Service-Oriented Architecture (SOA) is a paradigm for modeling the interaction of different parties in a distributed system. In SOA, a high abstraction level leads to platform-independent interoperability. Moreover, different parties are only loosely coupled to each other. As a result of these, SOA is a scalable and flexible architecture.

As industrial automation systems are typically inflexible and expensive to install or to modify, it would be beneficial to have all devices interact in the SOA manner. However, current technologies to implement a SOA are problematic from the devices point of view. The technologies require a lot of computational resources, and they also lack support for hard real-time functions. Work has been done to overcome these challenges, but especially hard real-time capable SOA cannot currently be implemented.

Despite their limitations, current SOA technologies can be used for several functions of industrial plants. In this study, service-oriented solutions are created for the estimation of environmental footprints and for condition monitoring. The solutions are modeled as diagrams using a standard graphical notation after which the diagrams are converted to an executable language.

Both implementations show the efficiency of the selected modeling method. The principles of SOA enable the reuse of different resources flexibly in different applications which saves work. A standard structured data format was used in both solutions, and it facilitates integration. As there is a built-in support for the format in modern applications, a solution designer can concentrate on data contents on a high level. Compatibility problems were also encountered, but they were overcome using wrapper services. There were also other integration problems with the technologies used. Despite the problems, graphical modeling saves time compared to textual methods to model communication. It was also recognized that careful design is required in distributed systems to avoid performance problems.