Data engineering approach to design of Web Services

With the wide acceptance of Web Services as the preferred implementation platform for service-oriented applications there is increased interest in how such applications should be designed. While there are similarities between software components and services there is now a general agreement that mapping existing components directly to Web Services leads to suboptimal design and results in poor performance and scalability. Most practitioners recommend the use of coarse-grained, message-oriented Web Service that minimize the number of messages and avoid the need to maintain state information between invocations. We argue that the design of message structures used as Web Services payloads directly impacts on application interoperability, and that excessive use of coarse-grained, document-centric message structures results in poor reuse and undesirable interdependencies between services. Our approach provides a framework for the design message structures using data engineering principles. We consider the impact of increasing message granularity on cohesion and coupling of service-oriented applications and analyze associated design trade-offs.