DOI: 10.36347/sjet.2018.v06i12.012

Pages: 454-463

The proliferation of artificial intelligence (AI) workloads in enterprise environments has necessitated sophisticated resource management strategies that balance performance, scalability, and cost efficiency. While containerization technologies such as Kubernetes have emerged as optimal platforms for deploying AI workflows, existing autoscaling mechanisms remain largely reactive and lack cost-awareness, leading to suboptimal resource utilization and unpredictable operational expenses in hybrid cloud environments. This paper presents a comprehensive framework for cost-aware autoscaling and resource prediction models specifically designed for AI workflows operating in hybrid Kubernetes–OpenStack clouds. Building upon established container orchestration principles, this research extends prior work by introducing predictive autoscaling mechanisms that leverage historical workload patterns, dynamic resource allocation strategies across bare-metal, virtual machine, and container layers, and cost-aware decision frameworks optimized for hybrid and private cloud deployments. The proposed framework integrates machine learning-based workload prediction models with multi-objective optimization algorithms to achieve simultaneous improvements in cost efficiency, resource utilization, and quality of service. Through comprehensive analysis of existing literature and synthesis of proven methodologies, this paper establishes theoretical foundations for next-generation autoscaling systems that can intelligently balance computational requirements with budgetary constraints. The research contributes three novel components: a predictive resource demand model for AI workloads, a cost-aware scheduling algorithm for hybrid infrastructure, and a dynamic allocation framework that optimizes across multiple infrastructure layers. These contributions address critical gaps in current container orchestration systems and provide actionable insights for organizations deploying AI workloads in cost-s