Title: Artificial Intelligence Across Disciplines: Research, Innovation, and Intelligent Solutions
Editors: Dr. Subita Bhagat, Dr. A. Balamurugan, Dr. P. Krishna Kumar, and Mrs. S. Nandhini Devi
ISBN: 978-81-69857-83-3
Chapter: 1
DOI: https://doi.org/10.59646/815/01
Author: S Rubini
Abstract
Modern hyperscale cloud data centers face escalating operational challenges driven by highly variable, bursty computational demands, strict Service Level Agreements (SLAs), and soaring energy consumption. Conventional threshold-based dynamic resource orchestration techniques struggle with reactive lag, causing SLA degradation or excessive over-provisioning. This chapter proposes a self-adaptive, intelligent resource provisioning framework that integrates deep predictive telemetry modeling with reinforcement learning to orchestrate compute, memory, and bandwidth allocations dynamically. Utilizing an edge-to-core industrial Internet of Things (IIoT) analytics platform as a representative case case study, our framework deploys a Temporal Convolutional Network (TCN) fused with Bidirectional Long Short-Term Memory (Bi-LSTM) networks for multi-horizon workload forecasting, paired with a Deep Deterministic Policy Gradient (DDPG) control agent for proactive virtual machine (VM) autoscaling and physical host consolidation. Rigorous empirical validation demonstrates a 31.4% reduction in overall data center energy expenditure, a 64.2% drop in SLA violation rates, and a 42.8% improvement in resource utilization compared to baseline heuristic schedulers.
Keywords: Intelligent Resource Management, Deep Reinforcement Learning, Cloud Workload Forecasting, Deep Deterministic Policy Gradient, Energy Optimization, Quality of Service.