SACSO: An Intelligent Algorithm for Energy Efficient Workload Scheduling in Cloud Computing

Abstract

Cloud computing provides scalable and on-demand computational resources, but the increasing energy consumption of cloud datacenters has made energy-efficient task scheduling an important research challenge. The task scheduling problem becomes more complex in heterogeneous cloud environments because minimizing execution time alone may increase the number of active resources and consequently increase energy consumption. This study proposes a Smooth Adaptive Cat Swarm Optimization (SACSO) algorithm for energy-efficient task scheduling in cloud computing. The proposed approach integrates a Smooth Adaptive Inertia Weight (SAIW) mechanism into the tracing mode of Cat Swarm Optimization (CSO) to achieve an effective balance between exploration and exploitation and to reduce the possibility of premature convergence. The scheduling model jointly considers makespan and energy consumption as the primary optimization objectives. The proposed SACSO is implemented in Java using the CloudSim 3.0 simulator and evaluated against CSO, LDCSO, Grey Wolf Optimization (GWO), and Particle Swarm Optimization (PSO). Experiments are conducted using three scientific workflow workloads, namely CyberShake, Inspiral, and Sipht, with 1000 tasks in each workload and heterogeneous virtual machines. Each algorithm is executed for 15 independent rounds to assess the stability of the obtained results. The experimental results demonstrate that SACSO consistently achieves lower makespan and energy consumption than the compared algorithms across all three workloads. Compared with LDCSO, SACSO reduces makespan by 11.66%–22.69% and energy consumption by 15.11%–20.69%. SACSO also achieves the highest throughput among the evaluated algorithms. Convergence analysis further demonstrates that SACSO reaches high-quality solutions rapidly during the initial iterations and maintains stable convergence during subsequent iterations. The results indicate that the proposed SAIW-based tracing mechanism effectively improves the exploration-exploitation balance of CSO and provides an efficient approach for energy-aware task scheduling in heterogeneous cloud environments.