Next-generation Cloud Computing: From Virtualization to AI-driven Cloud Services

Abstract

Cloud computing is evolving from infrastructure virtualization toward intelligent, automated, and adaptive service management. Although conventional cloud platforms provide elastic access to computing, storage, networking, and software resources, highly variable workloads create persistent challenges in resource allocation, autoscaling, service-level agreement (SLA) compliance, operational cost, and energy efficiency. This paper presents a structured AI-driven cloud resource management framework that integrates monitoring and AIOps, workload prediction, intelligent resource allocation, dynamic orchestration, and closed-loop feedback. The framework is complemented by a task-oriented taxonomy that maps artificial intelligence techniques to cloud-management functions and measurable objectives. A critical review of recent studies on deep reinforcement learning for scheduling, machine-learning-based autoscaling, adaptive resource prediction, load balancing, and energy-aware management is used to identify gaps in fragmented decision-making and inconsistent evaluation. The proposed architecture is designed to support virtual machines, containers, serverless services, edge–cloud deployments, and multi-cloud environments. A reproducible evaluation methodology is defined using response time, throughput, resource utilization, cost, energy consumption, prediction error, and SLA violations. The paper does not claim unmeasured performance improvements; instead, it provides an implementable framework and validation protocol that can be experimentally tested using public workload traces or a controlled cloud testbed.