Performance Analysis of CPU Utilization in Cloud Environments using Machine Learning Models

Cloud environments offer extensive services, making them a preferred choice for businesses and startups to host and maintain databases. Efficient resource management, including automatic scaling, is crucial for sustaining virtual machine environments. This paper addresses the challenge of forecasting CPU workloads to optimize resource allocation. Terraform was utilized to create and launch instances in the cloud environment, specifically using Amazon Web Services (AWS). A CPU-intensive application was executed on the server, and CPU utilization data was collected at regular intervals to serve as the dataset for the experiment. Three models—Gated Recurrent Unit, Long Short-Term Memory Autoencoder, and Seasonal Auto-Regressive Integrated Moving Average—were compared for short-term workload prediction. The results demonstrate that SARIMA outperforms the LSTM autoencoder by 2.91%, GRU by 1.78%, and LSTM autoencoder by 1.13% based on MAPE, MSE, MAE, and RMSE metrics.