The Service-Oriented Computing (SOC) promotes assembling software components into loosely-coupled networks of services, to create flexible, agile applications and business processes that span organizations and computing platforms. Due to the distributed and asynchronous nature of Web services, the Web service composition process plays an important role in achieving service-oriented computing. In dynamic environments by which contextual information such as Web service properties, business requirements, software and hardware capabilities often change, the composition process at design time and runtime is thus affected and, consequently, should be able to adapt composite applications or business processes to changes. Unfortunately, current Service-Oriented Architecture (SOA) approaches and automated Web service composition mechanisms fail to deal with continuous and unpredictable changes since they consider that the design and the runtime environments are stable and static. Building resilient service-oriented architectures that are adaptable to endogenous and exogenous changes in dynamic environments reveal a drastic challenge to current composition processes. In addition, current composition processes provide a limited support for business users to specify their requirements in business languages in order to semi- or automatically compose business processes (i.e., composite services). By such, the gap between business requirements and Web service related composition requirements increase the complexity of developing adaptable SOA-based applications or processes in dynamic environments.
To overcome these challenges, we introduce the concept of Resilient Service-Oriented Computing (rSOA) to construct resilient SOA-based applications and business processes driven by business requirements in dynamic environments. To this end, the resilient SOA is designed as a set of models that affect and are affected by each other. In particularly, the Web service composition model is viewed as the cornerstone in our model-to-model transformation approach. In this thesis, our contributions, which particularly focus on the requirement model and the composition model, consist of three main parts: firstly a multi-level requirement model provides business users with a structured natural-based language to specify composition requirements; secondly, a two-phase requirement transformation process derives an rule-driven Web service composition requirement model as set of multi-objective constraints on various variables related to the composition process and its dynamic environment. Each rule is a constraint that may represent any information, event, variation or change that may occur; thirdly, an ad-hoc Web service composition algorithm is introduced to flexibly construct composite services without predefined composition plans in dynamic environments while satisfying all constraints derived from users’ business requirements. The rule as a fine-grained element in the rule-driven Web service composition requirement model may also affect other model, such as business model or fault tolerance model, and may recursively invoke the model-to-model transformation approach to replan the ad-hoc Web service composition algorithm.