While the requirements of enterprise and web applications have driven the development of Cloud computing, some of its key features, such as customized environments and rapid elasticity, could also benefit scientific applications. However, neither virtualization techniques nor Cloud-like access to resources is common in scientific computing centers due to the negative perception of the impact that virtualization techniques introduce. In this paper we discuss the feasibility of the IaaS cloud model to satisfy some of the computational science requirements and the main drawbacks that need to be addressed by cloud resource providers so that the maximum benefit can be obtained from a given cloud infrastructure.