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: 14
DOI: https://doi.org/10.59646/815/14
Authors: Dr. M. Senthilkumar, and Dr. A. Balamurugan
Abstract
Modern hyperscale cloud data centers require adaptive orchestration frameworks capable of mitigating non-linear workload fluctuations while minimizing operational expenditure, carbon footprints, and Service Level Agreement (SLA) violations. This chapter presents an end-to-end artificial intelligence-driven resource allocation framework tailored for heterogeneous cloud infrastructures hosting high-concurrency microservices and mission-critical batch workflows. By coupling spatio-temporal workload forecasting via bidirectional temporal transformers with a multi-objective deep reinforcement learning scheduler based on proximal policy optimization, the system dynamically provisions virtualized compute, memory, and network resources. Evaluated over 30 days of production telemetry across thousands of virtual machines, the proposed model balances power usage effectiveness, thermal dissipation, and tail latencies. The empirical results demonstrate a 26.4% reduction in overall energy consumption and an 81.2% drop in SLA degradation penalties compared to conventional heuristic methods, proving the viability of autonomous, closed-loop machine learning for sustainable, resilient enterprise cloud infrastructure.
Keywords: Cloud Computing, Dynamic Resource Allocation, Deep Reinforcement Learning, Workload Forecasting, Energy Efficiency, Service Level Agreements.