A hybrid RL–GA–LSTM–AE framework for energy-aware and SLA-driven task scheduling in cloud computing environments

The rapid advance of cloud computing, along with the growing number of cloud services, has driven a rising demand for intelligent task-scheduling methods to enhance energy efficiency while maintaining SLA (Service Level Agreement) compliance. We present a Reinforcement Learning Genetic Algorithm framework, enhanced with a Long Short-Term Memory Autoencoder, to create a more adaptive and predictive scheduling mechanism in a dynamic cloud setting. Concretely, the LSTM Autoencoder extraction module is used to predict overload scenarios and detect abnormal behaviour, while the RL-GA optimises task allocation via multi-objective evolutionary learning. We validate our model through extensive simulations in a CloudSim environment using Google Cluster Traces and RUBiS datasets and demonstrate that, compared with traditional schedulers, the proposed model is much more efficient. The framework achieves a 27.5% reduction in makespan over baseline methods, reduces SLA violations by up to 4.2% (67% improvement), reduces deadline misses by 78.6%, improves CPU utilisation to 88.7%, and reduces static provisioning energy usage by up to 42.7%. ANOVA testing (with p-values < 0.001) supports the robustness of these results, and convergence analysis shows stable learning across all components of the framework within 500 training cycles. It confirms that it is responsive enough to support dynamic cloud environments, achieving 156.8 milliseconds of end-to-end scheduling latency in a real-time performance evaluation. Its near-linear computational complexity is validated through scalability testing with workload sizes ranging from 100 to 1000 tasks. These results and the comparison show the general effectiveness, scalability, and sustainability of the proposed RL-GA-LSTM-AE framework for intelligent resource scheduling in large-scale cloud data centres, and open the door for further developments in the design and realisation of green, SLA-driven computing infrastructures towards next-gen computing.