Based on system dynamics modeling method, the nonlinear model of the cloud computing virtual machines scheduling is built. With system dynamics methodology and the use of simulation tools, from a system perspective on cloud computing service requirements and building scale, the key factors affecting the number of cloud computing machines are identified. Furthermore, mathematical statistical theory and methods are used to find out optimal estimation and implementation of the construction scale with regarding to the decisive factor. The regression equations are submitted and the credibility of the regression equations is verified.