SARIMA Techniques for Predictive Resource Provisioning in Cloud Environments
Abstract
Seasonal Autoregressive Integrated Moving Average (SARIMA) models for dynamic cloud resource provisioning are introduced and evaluated in this work. Various cloud-based apps provided historical data to train and evaluate SARIMA models. The SARIMA(1,1,1)(0,1,1)12 model has an MAE of 0.056 and an RMSE of 0.082, indicating excellent prediction ability. This model projected resource needs better than other SARIMA settings. Sample prediction vs. real study showed close congruence between projected and observed resource consumption. MAE improved with hyperparameter adjustment, according to sensitivity analysis. Moreover, SARIMA-based resource allocation improved CPU usage by 12.5%, RAM utilization by 20%, and storage utilization by 21.4%. These data demonstrate SARIMA's ability to forecast cloud resource needs. SARIMA-based resource management might change dynamic cloud resource management systems due to cost reductions and resource usage efficiency. This research helps industry practitioners improve cloud-based service performance and cost.