In recent years, Cloud Computing has become popular to facilitate the use of Modeling and Simulation (M&S) resources. Nevertheless, there are still various issues to solve, including the structure constrain of chosen web service frameworks, the sharing of varied resources in the Cloud, and the difficulties in reproducing experiments. We show a new architecture based on Cloud Computing and new modeling methods to deal with these issues. This layered architecture, called Cloud Architecture for Modeling and Simulation as a Service (CAMSaaS), simplifies the deployment of M&S resources as services in the Cloud. CAMSaaS supports hierarchical resource services, experimental frameworks, scalable infrastructure and makes everything as a service. We deploy varied M&S resources as services in the Cloud, and build a Modeling and Simulation as a Service (MSaaS) middleware called CloudRISE to manage a variety of M&S resources. We also use the experimental framework concept to simplify the management of experiment environments. We present a case study for crowd evacuation application using the architecture.