Workload-Aware Energy-Efficient Query Scheduling for Cloud Database Systems: Experimental Study

Abstract

Energy efficiency is a growing challenge in cloud database systems, particularly for analytical workloads with intensive CPU and disk I/O demands. Traditional query scheduling strategies, such as First-Come First-Served (FCFS) and Shortest Job First (SJF) focus on performance optimization and do not explicitly consider energy consumption. This paper proposes an Energy-Driven Adaptive Scheduling (EDAS) strategy that prioritizes queries based on estimated CPU and disk I/O costs without modifying the database engine. Experiments were conducted on a cloud-based MySQL system using light, medium, and heavy workloads derived from Sakila and TPC-H benchmarks. Results showed that energy-aware scheduling is workload-dependent: SJF performed well under light and medium workloads, while EDAS achieved measurable energy savings under heavy workloads and greater resilience under CPU throttling. The study demonstrates the importance of workload-aware query scheduling for improving cloud database energy efficiency.