Increasingly, distributed systems are being used to provide enterprise level solutions with high scalability and fault tolerance These solutins are often built using Web servces that are composed to perform useful business functions Acceptance of these composed systems is often constrained by a number of non-functional properties of the system such as availability, scalability and performance There are a number of drstribution patterns that each exhibit different non-functional charactmstics These patterns are re-occuring distribution schemes that express how a system is to be assembled and subsequently deployed. Traditional approaches to development of Web service compositions exhibit a number of Issues Firstly, Web service composition development is often ad-hoc and requires considerable low level coding effort for realisatlon Such systems often exhibit fixed architectures, making maintenance difficult and error prone Additionally, a number of the non-funchonal reqwements cannot be easily assessed by exammng low level code. In this thesis we explicitly model the compositional aspects of Web service compositions usmg UML Activity diagrams Ths approach uses a modehng and transformation framework, based on Model Dnven Software Development (MDSD), going from high level models to an executable system The framework is guided by a methodological framework whose primary artifact is a distribution pattern model, chosen from the supplied catalog. Our modelling and transfomation framework improves the development process of Web service compositions, with respect to a number of criteria, when compared to the traditional handcrafted approach Specifically, we negate the coding effort traditionally associated with Web service composition development Maintenance overheads of the solution are also slgnificantly reduced, while improved mutability 1s achieved through a flexible architecture when compared with existing tools We also improve the product output from the development process by exposing the non-functional runtime properties of Web service compositlons using distribution patterns.