Improved Ensemble Learning-based VM Migration Prediction Technique

Virtual Machine computing resources significantly impact cloud storage performance and, in this regard, need to be well managed to reduce operational expenses. The scheme for predicting VM migration proposed in this paper is based on the ensemble learning technique with CPU usage as the primary performance parameter. We also assess the XGBoost, Random Forest, Gradient Boosting, and K-Nearest Neighbors algorithms and compare the Stacking, Voting, and Blending ensemble approaches. We use the real-world VM time-series performance data collected from the GWA-T-13 Materna dataset for empirical validation. These findings show that ensemble approaches achieve higher accuracy and generalization when compared with the individual model. However, this work is helpful in understanding strategies for moving virtual machines and getting some advice for resource management in cloud environments.