Cloud computing allows to build sophisticated software sys-tems on virtualized infrastructures at a fraction of the cost that was necessary just a few years ago. The deployment of such complex systems, though, is still a serious issue due to the need of deploying a large number of packages and services, their elaborated interdependencies, and the need to define the (ideally optimal) allocation of software compo-nents onto available computing resources. In this paper we present the Aeolus Blender (Blender in the following), a toolchain that automates the assembly and deployment of complex component-based software systems in the "cloud". By relying on a configuration optimizer and a deployment planner, Blender fully automates the deploy-ment of real-life cloud applications on OpenStack infrastruc-tures, by exploiting a knowledge base of software compo-nents defined in the Mandriva Armonic tool-suite. The final deployment is guaranteed to satisfy not only user require-ments and software dependencies, but also to be optimal with respect to the number of used virtual machines.