A Chemistry-Inspired Middleware for Self-Adaptive Service Orchestration and Choreography

Orchestration and choreography are two perspectives to model service interactions in executing a service composition. Compared to orchestration model, choreography can improve the performances in scalability, throughput and execution time by eliminating centralized point of coordination. However, it also brings additional complexities and challenges such as inter-organizational service coordination, fault-tolerance and etc. In this paper, we propose two self-managed and self-adaptive choreography models, namely semi-choreography and auto-choreography. Both models as well as a centralized orchestration model have been implemented in a chemistry-inspired middleware. With the metaphor from chemical reactions, the middleware is modeled as an autonomic distributed chemical system, where services are described as complex molecules and service coordination and adaptation is performed by a series of pervasive chemical reactions controlled by a number of rules. By using different sets of rules, providers are able to specify their preferred execution models to run service compositions, in either centralized or collaborative way. The middleware is implemented in Higher-Order Chemical Language (HOCL) and running on distributed infrastructures (Grid'5000). Finally, by executing two experimental workflows in the middleware, a number of experiments have been conducted to evaluate and compare different models in terms of complexity and efficiency.