Ayush Raj Jha, Milan Parikh, Aniket Abhishek Soni and Sneja Mitinbhai Shah
Cloud data centers face increasing pressure to reduce operational energy consump- tion as big data workloads continue to grow in scale and complexity. This paper presents a workload-aware scheduling framework that uses profiling of CPU usage, memory demand, and storage I/O behavior to guide energy- efficient virtual machine (VM) placement. By combining historical execution logs with real- time telemetry, the system predicts the en- ergy and performance impact of candidate placement decisions and adaptively consolidates workloads without violating service-level agree- ments (SLAs). The framework was evaluated using representative Hadoop MapReduce, Spark MLlib, and ETL workloads on a multi-node cloud testbed. Experimental results demonstrate a consistent reduction of 15–20% in energy consumption while maintaining SLA compli- ance. These findings highlight the effectiveness of data-driven workload profiling as a practi- cal strategy for improving the sustainability of cloud computing environments.