Deep Reinforcement Learning for Intelligent Task Scheduling in Heterogeneous Cloud Environments

This study proposes an efficient task scheduling strategy in cloud computing environments using a Deep Reinforcement Learning (DRL) algorithm to optimize system performance under dynamic workloads. The approach is evaluated using the HPC2N workload within the CloudSim simulation platform and benchmarked against existing algorithms, including ACO, RAPTS, and HDDPGTS. The proposed DRL-based scheduler exhibits superior adaptability and intelligent decision-making capabilities by learning optimal task-to-resource mappings through continuous interaction with the cloud environment. Simulation results demonstrate that the DRL method achieves a $32.4 \%$ reduction in makespan, $48.7 \%$ lower energy consumption, and $22.6 \%$ improvement in resource utilization on average compared to the baseline techniques. These improvements highlight the DRL algorithm’s effectiveness in achieving a balanced trade-off between performance and energy efficiency, making it a promising solution for real-time and large-scale cloud computing applications.