Pattern-based Deployment Models Revisited: Automated Pattern-driven Deployment Configuration

The manual deployment of cloud applications is error-prone and requires significant expertise. Therefore, many deployment automation technologies have been developed that enable deploying applications fully automatically by processing deployment models. However, while these technologies substantially simplify deployment, the manual creation of deployment models ironically poses similar challenges to manually deploying applications as technical expertise about the components to be deployed and their dependencies is required. Therefore, we introduced Pattern-based Deployment Models (PbDMs) in a previous work that allow using design patterns to model components in an abstract manner, which are then automatically replaced by concrete technologies. However, in many scenarios, the resulting deployment models still have to be subsequently adapted with regard to the configuration of the selected technologies, e. g., to configure a selected Platform as a Service (PaaS) offering, such as Amazon Beanstalk, for optimal scaling. Therefore, while our previous work only enables using design patterns to model components, in this paper we extend the proposed meta-model and algorithms by the possibility to specify behavioral aspects of components and relations also in the form of patterns. Moreover, we show how these annotated patterns can be automatically transformed into concrete configurations that reflect their semantics. We present a prototype and a case study to validate the extension’s practical feasibility. Keywords-Deployment Automation; Deployment Modeling; Patterns; Model-driven Architecture; TOSCA.