AbstractThe Hybrid Cloud Infrastructure has been adopted as a paradigm of modern enterprise computing that unites the scalability of a public cloud service environment with the security and control of a chosen on-premises enterprise resources. Resource provisioning in such environments that are part hybrid is essential in ensuring the best performance, cost-effectiveness, and service-level agreement (SLA) are met. The paper provides a detailed study on the resource provisioning strategies such as the static, dynamic, reactive, and proactive methods of hybrid cloud systems. The predictive provisioning, workload-aware scheduling, and cost-optimization frameworks, which are machine-learned, are assessed in the context of simulation analysis. We show that when used in proactive provisioning, ML significantly reduces resource over-provisioning (by 34 percent), minimizes the average response latency (by 28 percent), and also is much cost-effective (up to 41 percent) as compared to traditional threshold-based strategies. Further challenges that we find to be open include federated resource orchestration, multi-tenant isolation, and green cloud provisioning. The work adds a single taxonomy of provisioning strategies and performance benchmarking framework of hybrid cloud environments.
R et al. (2026) studied this question.