A Comprehensive Review of Load Balancing Optimization in Cloud Infrastructure: From Metaheuristic Algorithms to Hybrid Deep Learning Approaches

ABSTRACT:

Scheduling algorithms often struggle to simultaneously optimize performance, energy efficiency, and quality of service (QoS) under rapidly changing demand patterns. Recent advances in bio-inspired metaheuristic optimization and deep learning have opened new research directions for intelligent resource management in large-scale cloud infrastructures. This review provides a comprehensive and structured analysis of cloud computing architectures, task scheduling mechanisms, load balancing techniques, and optimization metrics, with particular emphasis on swarm intelligence approaches such as Ant Colony Optimization (ACO), Bat Optimization (BO), and Particle Swarm Optimization (PSO), as well as deep learning-based scheduling strategies. By critically synthesizing existing literature, this paper identifies key limitations in current approaches, including poor adaptability to bursty workloads, limited energy-awareness, and insufficient integration of predictive intelligence. Motivated by these gaps, the review positions a hybrid optimization framework that integrates Hybrid Bat Colony Optimization (HBCO) for initial resource placement, Ant Colony Optimization (ACO) for dynamic load redistribution, and Deep Learning (DL) models for long-term workload prediction and policy refinement. The proposed direction aligns with emerging trends in intelligent, green, and autonomous cloud resource management, providing a strong foundation for future research and practical deployment.

Keywords: Cloud computing; Task scheduling; Load balancing; Ant Colony Optimization; Bat Algorithm; Deep learning; Resource utilization.