Cost-Optimized Cloud Scheduling for ETL and Big Data Using AI
Chaitanya Krishnama, Raghavender Puchhakayala, Sudarshan Kotha, Faiz Gouri · 2025
Cloud-based data pipelines are critical for large-scale ETL and big data analytics, yet in-efficient scheduling leads to high costs and resource underutilization. Traditional approaches, such as static provisioning and rule-based auto-scaling, fail to adapt to dynamic workloads and ever-changing cloud pricing. This paper introduces an AI-driven cost-aware scheduling framework that integrates LSTM-based workload forecasting, Isolation Forest anomaly detection, and real-time cost analytics. Unlike purely reactive or threshold-based methods, this approach proactively adjusts resources to minimize costs and prevent performance bottlenecks. Experiments using the Google Cluster Workload Traces dataset and Azure VM Pricing data demonstrate up to a 30-40% reduction in operational expenses, along with improved processing efficiency and auto-scaling responsiveness. However, the proposed solution may require further adaptation for extreme-scale scenarios or specialized compliance environments. These findings highlight the novelty of combining predictive analytics, anomaly detection, and cost optimization, offering a robust strategy for enhancing performance and reducing expenses in cloud-based ETL and big data workflows.