Article Preview
Top1. Introduction
Today many institutions in almost every sector strive to improve their mutual collaboration, increase data sharing, and decrease the response time in order to provide high-quality services to their customers. Service-oriented architectures (SOA) based on Web services are seen as an essential platform to fulfill this goal. A good progress in several areas of SOA has been made and several standards for web services cover the most important aspects of software services. However, developing applications for SOAs represents a major shift for the developers. The long term vision for SOA-based systems lies in open environments where software services are not a-priori known to the system developers. On the one hand, the system development can fully focus on the business process design aspects and leave binding to concrete services for the later stages of the development. On the other hand, this approach needs to be supported by the mechanisms that allow for capturing design decisions at the business process levels to be enforced and utilized at the implementation level. In reverse, services to be selected and properly integrated into the compositions implementing the business process must be better described.
The enthusiasm for SOAs and service-oriented systems is growing with the increasing number of standards for these technologies. Many basic aspects of SOAs and Web services have been standardized enabling increase in implementation and deployment of SOA-based systems. However, as we demonstrate in the literature review section below, the standards are fragmented and a major effort by software developers is needed to bridge between them. The efficient development of high-quality systems based on SOAs requires extension of standards, modeling languages, and system development tools to facilitate system development process.
In this paper, we investigate some key research directions in addressing the above mentioned problems. Our primary focus is on investigating the promises of a novel software engineering area – Model-Driven Engineering (MDE) – for the development of service-oriented systems. MDE offers modeling methods that increase the level of abstraction when developing software, shifting from low-level platform specific details towards high-level software modeling approaches (i.e., computation-independent and platform-independent models). This not only allows us to reuse the same software design on various implementation platforms but more importantly to build more reliable service-oriented solutions which better responds to both stakeholder requirements and business rules and which is verified in early design stages. In Appendix A, we provide a brief introduction into the terminology of MDE.
Presently, development of SOAs and service-oriented systems mainly follows proven principles of business process modeling. Based on process-oriented modeling languages (e.g., Business Process Modeling–BPMN notation and UML Activity Diagrams), service developers start from the high-level business process models. Such models typically cover aspects such as flow of control, data, and activities. Once created, business process models are used as either blueprints for construction of service-oriented systems or even more ambitiously they are automatically translated into (skeletons of) implementation artifacts such as executable service compositions in Business Process Execution Language (OASIS BPEL, 2007). A significant progress in this area has been made in several critical aspects (see Sect. 2 for some details). In spite of the promising results, there are several research challenges to be addressed. In this paper, we outline some research challenges: