Managing CPU resource is essential for maintaining performance in edge computing environments. However, it is rather difficult to deal with ever-changing workload by using the conventional methods, Static CPU Throttling or standalone application of Dynamic Frequency Scaling (DFS). This leads to unnecessary performance degradation and power consumption, which pose a significant problem. In order to address this issue, this paper introduces an optimization approach to efficiently control CPU usage by combined methods of Model Predictive Control (MPC) and Dynamic Frequency Scaling (DFS). According to the experimental results, MPC makes it possible to manage CPU usage and the integration of MPC and DFS leads to a maximized optimization effect. Furthermore, it is confirmed that the proposed method provides a statistically significant performance improvement through the T-test.