Cloud computing, today, is a ubiquitous paradigm. Its features such as availability of a practically infinite pool of computing resources, on demand, by using a pay-per-use model has resulted in its adoption by the industry for the realization of modern, sophisticated, and highly scalable IT applications. Such applications are often comprised of various components and services offered by different cloud service providers. This, in turn, raises two significant challenges- (i) automated provisioning and management, and (ii) interoperability and portability of the applications in a multi-cloud environment. As a result, the Topology and Orchestration Specification for Cloud Applications (TOSCA) standard was introduced by OASIS. This standard provides a metamodel to describe the topology of complex applications along with all the components, artifacts, and services in a single template that allows deploying the application in an interoperable and portable manner. In this Master thesis, we propose a concept that generates small and reusable TOSCA provisioning plans which can be orchestrated to deploy the overall application as opposed to using a monolithic provisioning plan. This goal is achieved in three steps - (i) splitting the application topology into a set of smaller sub-topologies, (ii) generating smaller plans called partial plans for each sub-topology, (iii) and finally orchestrating the partial plans to provision an instance of the application. Additionally, this concept enables the reuse of these plans for tasks such as scaling out individual components of the application. Finally, the feasibility of the proposed concept is demonstrated by a prototypical implementation developed using the OpenTOSCA framework.