Predictive Resource Orchestration for AI-Driven Healthcare Workloads in Multi-Data-Centre Cloud Migrations

Workload demands in healthcare are being significantly influenced by artificial intelligence (AI) and require substantial computing power, fast processing speed (low latency), as well as consistent and high-quality operation at dispersed locations around the globe through geographically distributed data centers. Cloud migration of multi-data-center systems presents major challenges related to the orchestration of resources due to variability of workload, heterogeneity of infrastructure, and the necessity for strict Service Level Agreements (SLAs). This paper presents a predictive resource orchestration framework that utilizes AI-based forecasted workloads to optimize the use of resources and scheduling of tasks as well as scaling of resources in the context of multiple cloud data centers. The predictive resource orchestration framework is made up of machine learning models used for predicting workloads and an adaptive scheduling algorithm for dynamic resource allocation based on forecasted demand for the minimization of latency, operating costs and SLA failures. Results from simulation studies indicate that the predictive resource orchestration framework provides better use of resources and shorter time to complete tasks than the heuristic or rule-based scheduling strategies that are commonly used today. The predictive resource orchestration framework represents a scalable method for the management of AI-driven health care workloads and provides both higher quality and higher reliability in the process of migrating applications between data centers of a multi-data-center cloud system.