Energy-Efficient Big Data Processing Using Adaptive Resource Scheduling in Cloud Environment

Nidhin Karunakaran Ponon

Abstract: This paper proposes an energy-efficient and SLA- aware task scheduling framework for large-scale cloud environments using a hybrid Genetic Algorithm?Whale Optimization Algorithm (GA-WOA) integrated with predictive modeling and explainable AI. The model uses XGBoost regressors to estimate task execution time and resource utilization based on five- dimensional workload vectors extracted from the Alibaba Cluster Trace v2018 dataset, which contains over 1.2 million real-world job instances. Scheduling is performed using a multi-objective fitness function that simultaneously minimizes total energy con- sumption, SLA violation rate, and makespan while ensuring task- to-resource exclusivity and capacity constraints. Experiments conducted on a CloudSim-based simulation environment with 200 physical hosts and 6 baseline methods demonstrated that the proposed approach achieves 93.4% precision, 92.1% recall, and 92.7% F1-score. Compared to the best baseline, the proposed model reduces SLA violations from 11.6% to 2.7%, energy usage from 152.6 kWh to 118.3 kWh, and load imbalance from 0.179 to 0.097. Root Mean Square Error (RMSE) was minimized to 0.131 for resource predictions. An ablation study confirmed the critical role of the prediction module, SLA constraint, and migration logic. SHAP-based explainability validated the model?s transparency by highlighting CPU demand and data size as dominant scheduling features.

Keywords: Energy-efficient scheduling, big data processing, cloud computing, resource optimization, SLA-aware migration, Alibaba Cluster Trace