This review article investigates the utilization of machine learning to optimize SAP workloads within complex multi-cloud enterprise environments. As global organizations move away from single-vendor dependence, the resulting architectural fragmentation introduces significant challenges regarding data gravity, network latency, and fluctuating egress costs. The study evaluates how predictive machine learning models, specifically Long Short-Term Memory networks and reinforcement learning agents, can be deployed to facilitate autonomous resource right-sizing and intelligent workload placement across hyperscalers like AWS, Azure, and Google Cloud. By leveraging the SAP Business Technology Platform and federated data architectures, enterprises can create a self-optimizing fabric that balances performance requirements with cost-efficiency. The research further examines the role of federated learning in maintaining data sovereignty and the technical hurdles of interoperability between disparate cloud APIs. Additionally, the paper explores emerging trends such as agentic AI for autonomous resource negotiation and sustainability-centric optimization to reduce the carbon footprint of data center operations. The article concludes that integrating machine learning into the orchestration layer is a strategic necessity for transforming SAP from a rigid, monolithic system into a liquid, cloud-agnostic platform capable of real-time adaptation to business demands.
Gayan Hettiarachchi (2024) studied this question.