A Review of Load Balancing Challenges in Fog Computing: Architectures, Strategies, and Emerging Trends

Abstract

Fog computing extends cloud distribution to the network edge, enabling low-latency processing for Internet of Things (IoT) applications. Load balancing in fogs is much more challenging than that in clouds because fogs are highly heterogeneous and distributed; devices may move from one region to another while employing services, and the locality of accessed services must be ensured based on latency. This work is a survey of the 2023-2025 period, reviewing 60 peer-reviewed papers and assessing the current state of load-balancing solutions. The survey identifies seven key challenge areas: resource heterogeneity management, latency-aware scheduling, energy-efficiency optimization, task offloading decisions, mobility management, fault tolerance, and security. Based on the surveyed literature, deep reinforcement learning accounts for the majority of work in this area, appearing in 66% of recent publications and improving response time by up to 84% relative to conventional alternatives. Furthermore, this work presents content such as a taxonomy of load balancing algorithms, comparative performance analysis via a variety of metrics, and also identifies important research gaps in the area, such as reliance on simulation-based approaches (more than 95% of the literature), lack of design for standardized benchmarking platforms, and insufficiency of embedding security properties in performance evaluation. We finally discuss future research directions in which federated learning, blockchain technology, and multi-agent reinforcement learning are integrated to build fog systems that are autonomous, secure, and adaptive.