Although building energy modeling has been common for many years, large-scale analyses have more recently become achievable for more users with access to affordable and vast computing power in the cloud. Recently, the OpenStudio Parametric Analysis Tool was enhanced, allowing users to launch their own cloud resources to run a large number of simulations. This paper discusses the process used to dynamically provision the machine images used in the cloud repeatedly and consistently, to launch and configure instances of these images, and finally to run analyses using the clustered instances running in the cloud. Several open source libraries for performing these tasks were used and extended as needed to support building energy modeling needs. This paper also describes various use cases of the OpenStudio Cloud implementation and outlines future use cases that are possible once a large clustered environment is available.