The success of cloud computing has radically changed the way in which services are implemented and deployed, and made accessible to external and remote users. The cloud computing paradigm, in fact, supports a vision of distributed IT where software services and applications are outsourced and used on a pay-as-you-go basis. In this context, the ability to guarantee an effective management of cloud performance and to support automatic scalability become fundamental requirements. Cloud users are increasingly interested in a transparent and coherent vision of cloud, where performance is guaranteed in different scenarios, and under different and heterogeneous loads. In this paper, we analyze the benefits of an integrated scalability approach at different layers of the cloud stack, focusing on the computing infrastructure and database layers. To this aim, we provide different performance metrics and a set of rules based on them to evaluate the status of the cloud stack and scale it on demand to maintain stable performance. We then implement a proof-of-concept architecture to experimentally analyze cloud performance in three scenarios of scalability: computing infrastructure only, database only, and the case in which computing infrastructure and database compete for resources.