<p>As Kubernetes is widely used for deploying and managing microservices, container orchestration tools are gradually becoming the standard for cloud-native environments. Kubernetes resource scheduling is a key technology in cloud computing. With the increase and diversification of business orders, the default Scheduler may face difficulties in acquiring resource requirements during peak hours. This can lead to node abnormalities. The default Scheduler relies on predefined configurations. This hinders optimal server resource allocation during peak periods. As a result, there is a significant disparity in resource usage between peak and off-peak times. To address this, this paper proposes a resource scheduling framework with monitoring and prediction modules. The monitoring module tracks real-time resource changes, while the prediction module forecasts resource demand. The scheduling process is optimized by predicting resource demands during peak periods. Experiments evaluate the prediction module using RMSE, MAE, and MAPE metrics. The results demonstrate that it outperforms the traditional Scheduler. In addition, experiments assesse the performance of the proposed framework in handling burst traffic. The results indicate that, compared to the default Scheduler, the framework achieves approximately 17% higher CPU usage and 8.5% higher throughput. This demonstrates a significant improvement in the container's ability to handle burst traffic. Compared to the original scheduling model, this model can effectively handle the growth in resource demands during peak periods. It ensures the stable operation of container scheduling.</p>