HCAS:A Hybrid Grey Wolf Optimizer–Deep Q-Network Scheduler for Carbon-Aware Sustainable Cloud Computing

The exponential growth of cloud data centers has elevated energy consumption and carbon emissions to first-class concerns alongside conventional performance objectives. Existing scheduling approaches either optimize traditional metrics such as makespan and response time without carbon awareness, or incorporate sustainability objectives only as secondary constraints with static estimates. This paper introduces HCAS (Hybrid Carbon-Aware Scheduler), a two-phase scheduling framework that couples the Grey Wolf Optimizer (GWO) for near-optimal initial task placement with a Deep Q-Network (DQN) reinforcement learning agent for dynamic, real-time workload rebalancing driven by real historical regional carbon intensity signals. The GWO phase minimizes a composite fitness function spanning makespan, energy, and carbon cost; the DQN phase continuously adapts to shifting utilization and historical carbon data obtained from Electricity Maps. HCAS is evaluated entirely within a simulation environment—CloudSim Plus—using real-world workload traces from the Google Cluster Trace 2019 dataset and historical carbon intensity data from Electricity Maps across three geographically distributed datacenter regions; validation on a physical multi-datacenter testbed remains an important direction for future work. Within this simulation scope, HCAS reduces carbon emissions by up to 34% and energy consumption by up to 28% compared to classical and standalone metaheuristic baselines, while simultaneously improving resource utilization and maintaining SLA violation rates below 2%. Ablation studies confirm that both phases contribute independently: GWO provides near-optimal initial placement that reduces the DQN’s exploration burden, while the DQN’s carbon-primary reward signal achieves emissions reductions that energy-only optimization cannot reach