It is a real challenge for cloud providers to offer cloud resources that are 'virtually unlimited and can be appropriated in any quantity at any time' in a cost-effective way. In this paper, we propose quantitative models that enable cloud providers to make an informed trade-off between cost and quality. Distinguishing between public and private cloud environments we consider infinite and finite source models respectively. In both cases either homogeneous or heterogeneous cloud resource requests are considered. These models can be applied to cloud dimensioning based on request blocking probability, an important SLA parameter. We derive a novel, insightful method that makes it possible to compute resource requirements in private clouds with heterogeneous resource requests. We show the importance of applying finite source models in the context of private clouds. We also use the proposed models to quantify the benefits of cloud federations.