HUNHODRL: Energy efficient resource distribution in a cloud environment using hybrid optimized deep reinforcement model with HunterPlus scheduler

Resource optimization and workload balancing in cloud computing environments necessitate efficient management of resources to minimize energy wastage and SLA (Service Level Agreement) violations. The existing scheduling techniques often face challenges with dynamic resource allocations and lead to inefficient job completion rates and container utilizations. Hence, this framework has been proposed to establish HUNHODRL, a newly-minted DRL-based framework that aims to improve container orchestration and workload allocation. The evaluation of this framework was done against HUNDRL, Bi-GGCN, and CNN methods comparatively under two sets of workloads with datasets on CPU, Memory, and Disk I/O utilization metrics. The model optimizes scheduling choices in HUNHODRL through a combination of destination host capacity vector and active job utilization matrix. The experimental results show that HUNHODRL outperforms existing models in container creation rate, job completion rate, SLA violation reduction, and energy efficiency. It facilitates increased container creation efficiency without increasing the energy costs of VM deployments. This method dynamically adapts itself and modifies the scheduling strategy to optimize performance amid varying workloads, thus establishing its scalability and robustness. A comparative analysis has demonstrated higher job completion rates against CNN, Bi-GGCNN, and HUNDRL, establishing the potential of DRL-based resource allocation. The significant gain in cloud resource utilization and energy-efficient task execution makes HUNHODRL and its suitable solution for next-generation cloud computing infrastructure.