Mahesh Loya,
Punam Shinde,
- Student, Department of Information Technology, Pune Institute of Computer Technology, Maharashtra, India
- Student, Department of Information Technology, Pune Institute of Computer Technology, Maharashtra, India
Abstract
Efficient load balancing remains one of the most critical challenges in cloud computing, particularly as modern cloud environments rely on heterogeneous virtual machines (VMs) with varying computational capabilities. The effective distribution of workloads across these resources directly influences system performance, energy efficiency, and service reliability. This paper presents a comparative evaluation of three major categories of load-balancing approaches: traditional scheduling heuristics, nature-inspired metaheuristic algorithms, and reinforcement learning (RL)-based techniques. The analysis is conducted across five key performance dimensions, including response latency, throughput, energy consumption, resource utilization, and fault resilience. A systematic review of ten peer-reviewed studies published between 2019 and 2024 was performed to identify current trends, strengths, and limitations of existing approaches. The findings indicate that RL-based schedulers have achieved significant improvements in simulated cloud environments, delivering energy savings of up to 56% and reducing response times by as much as 46% when compared with optimized conventional methods. However, these benefits are accompanied by substantial computational overhead during training, often requiring GPU-accelerated infrastructure. Additionally, RL models face scalability challenges due to state-space explosion when the number of virtual machines increases beyond several hundred nodes. The review further reveals that hybrid approaches, which combine the exploration capabilities of metaheuristic algorithms with the adaptive decision-making strengths of reinforcement learning, consistently produce superior performance across multiple evaluation metrics. Despite recent progress, several important research gaps remain unresolved. These include the lack of standardized benchmarking frameworks, limited integration of energy-aware scheduling objectives, inadequate fault-recovery mechanisms, and poor scalability of state representations. To address these challenges, this paper proposes a future research agenda focused on lightweight, scalable, and transfer-learning-enabled reinforcement learning agents capable of supporting next-generation cloud computing environments more efficiently and reliably.
Keywords: Cloud computing, load balancing, reinforcement learning, metaheuristic optimization, energy-aware scheduling, virtual machine allocation
[This article belongs to Journal of Operating Systems Development & Trends ]
References
- Kruekaew B, Kimpan W. Multi-objective task scheduling optimization for load balancing in cloud computing environment using hybrid artificial bee colony algorithm with reinforcement learning. IEEE Access. 2022;10:17803-18.
- Gures E, Shayea I, Ergen M, Azmi MH, El-Saleh AA. Machine learning-based load balancing algorithms in future heterogeneous networks: a survey. IEEE Access. 2022;10:37689-717.
- Gamal M, Rizk R, Mahdi H, Elnaghi BE. Osmotic bio-inspired load balancing algorithm in cloud computing. IEEE Access. 2019;7:42735-44.
- Souravlas S, Anastasiadou SD, Tantalaki N, Katsavounis S. A fair, dynamic load balanced task distribution strategy for heterogeneous cloud platforms based on Markov process modeling. IEEE Access. 2022;10:26149-62.
- Gao R, Xie X, Guo Q. K-TAHP: a Kubernetes load balancing strategy based on TOPSIS+AHP. IEEE Access. 2023;11:102132-9.
- Ramya K, Ayothi S. Hybrid prairie dog and beluga whale optimization algorithm for multi-objective load balanced-task scheduling in cloud computing environments. China Commun. 2024;21(7):307-24.
- Lahande PV, Kaveri PR, Saini JR, Kotecha K, Alfarhood S. Reinforcement learning approach for optimizing cloud resource utilization with load balancing. IEEE Access. 2023;11:127567-77.
- Mohammadian V, Navimipour NJ, Hosseinzadeh M, Darwesh A. Fault-tolerant load balancing in cloud computing: a systematic literature review. IEEE Access. 2022;10:12714-31.
- Jayanetti A, Halgamuge S, Buyya R. Deep reinforcement learning for energy and time optimized scheduling of precedence-constrained tasks in edge-cloud computing environments. Future Gener Comput Syst. 2023;137:14-30.
- Jayanetti A, Halgamuge S, Buyya R. Reinforcement learning based workflow scheduling in cloud and edge computing environments: a taxonomy, review and future directions [Preprint]. 2024. arXiv:2408.02938.
Journal of Operating Systems Development & Trends
| Volume | 13 | |
| Issue | 02 | |
| Received | 03/04/2026 | |
| Accepted | 03/07/2026 | |
| Published | 10/08/2026 | |
| Publication Time | 129 Days |