Towards The Deployment of Cloud Applications Using a DSL

Cloud infrastructures offer facilities to develop and deploy large-scale applications. Nevertheless, testing Cloud applications is an intricate task because applications themselves are very complex, infrastructures are highly distributed and heterogeneous, the amounts of data are massive, and the interconnections between services/components are sophisticated. From a business point of view, testing is generally not seen as an interesting phase of software development because, unlike the development of new features, it does not offer a direct return on investment. We think that providing accurate methodologies and tools to test Cloud applications will allow companies to speed up testing (and thus save time and money), and produce better software (and thus increase their competitiveness). In this research, we aim to ease the deployment of Cloud applications. This task is painful in the context of Cloud applications because testers have to write different scripts to test their applications with different Cloud vendors (e.g. Amazon EC21, Google Compute Engine 2). Our objective is to define a generic language that allows writing deployment scripts disregarding Cloud vendors. State of the Art Many different configuration management tools have been developed to maintain applications across multiple hosts in a distributed environment e.g., Cfengine (Burgess, 1995), Puppet (Kanies, 2006), and Chef3. A similar tool described by Juve et. al. (Juve, 2011; Juve, 2011), Wrangler, enables users to define application’s layout using a declarative language, and provides a robust service to automatically provision, configure and manage the application. Such configuration tools are often