Adaptive Bare-Metal Data Center Load Distribution Using Dual-Policy Deep Reinforcement Learning with Hysteresis Switching and Safety Shielding

Authors

  • Aay Ramdan Informatics, Telkom University, Indonesia
  • Muhammad Faris Fathoni Informatics, Telkom University, Indonesia
  • Setyorini Informatics, Telkom University, Indonesia

DOI:

https://doi.org/10.52436/1.jutif.2026.7.5.5844

Keywords:

Advantage Actor-Critic, Bare-Metal Server, Deep Reinforcement Learning, Load Balancing, Proximal Policy Optimization, Quality Of Service

Abstract

Load balancing in bare-metal server environments presents a unique challenge, as conventional static algorithms and single-agent Deep Reinforcement Learning (DRL) often fail to adapt securely to extreme traffic variations. Single DRL agents face a convergence dilemma: Advantage Actor-Critic (A2C) suffers from policy degradation under sudden stress, while Proximal Policy Optimization (PPO) lacks rapid responsiveness in low-traffic states. To address these limitations, this study proposes a Hybrid Adaptive Load Balancing (HALB) architecture that dynamically integrates A2C and PPO policies using hysteresis switching logic, fortified by a deterministic Safety Shield. The system was evaluated using an Nginx reverse proxy against standalone A2C and PPO baselines under Ramp-up, Sudden Spike, and High Traffic scenarios. Experimental results demonstrate that HALB provides noticeable improvements over single-policy agents, particularly in latency optimization and tail latency truncation. During extreme sudden traffic spikes, the HALB system constrained the 95th Percentile (P95) latency to 77.33 ms, compared to standalone A2C (85.00 ms) and PPO (85.23 ms). Under sustained high traffic conditions, HALB achieved an overall average latency of 23.64 ms while maintaining a load distribution fairness index (J > 0.99). Fortified by the Safety Shield that prevented node crashes at an 85% CPU threshold, the HALB architecture offers a balanced approach between intelligent responsiveness and operational stability. This research provides a resilient traffic management framework for rigid infrastructures, contributing to the development of safe adaptive control in non-elastic distributed systems.

Downloads

Download data is not yet available.

References

J. Kaur and S. Kinger, “A Survey on Load Balancing Techniques in Cloud Computing,” International Journal of Science and Research (IJSR)International Journal of Science and Research (IJSR), vol. 3, no. 5, pp. 1155–1159, 2014.

J. Jimenez et al., “Resource Allocation of Multi-User Workloads in Cloud and Edge Data-Centers Using Reinforcement Learning,” in 2023 19th International Conference on Network and Service Management, CNSM 2023, Institute of Electrical and Electronics Engineers Inc., 2023. doi: 10.23919/CNSM59352.2023.10327797.

T. Zhang, Q. Zhang, Y. Lei, S. Zou, J. Huang, and F. Li, “Load balancing with traffic isolation in data center networks,” Future Generation Computer Systems, vol. 127, pp. 126–141, Feb. 2022, doi: 10.1016/j.future.2021.09.002.

N. Vijay Kumar and P. Pattnaik, “A Dynamic Workload Prediction and Distribution in Cloud Computing using Deep Reinforcement Learning and LSTM,” Journal of Harbin Institute of Technology, 2024, doi: 10.11916/j.issn.1005-9113.2024053.

R. Z. Al Shaikh, M. M. Jawad Al-Nayar, and A. M. Hasan, “Reinforcement Learning Algorithms for Adaptive Load Balancing in Publish/Subscribe Systems: PPO, UCB, and Epsilon-Greedy Approaches,” Informatica (Slovenia), vol. 49, no. 7, pp. 77–88, 2025, doi: 10.31449/inf.v49i7.6895.

R. S. Sutton and A. G. Barto, Reinforcement Learning: An Introduction Second edition, in progress. Cambridge, MA, USA: MIT Press, 2018.

A. A. Ali, G. A. Hussein, B. M. Muter, and O. A. Hassen, “DeepBalance: A Deep Reinforcement Learning Framework for Dynamic Load Balancing in Software-Defined Networks,” Journal of Intelligent Systems and Internet of Things, vol. 17, no. 1, pp. 279–290, 2025, doi: 10.54216/JISIoT.170120.

Z. Wang, “Deep Reinforcement Learning for Resource Management in IoT-Edge-Cloud Computing Continuum: A Taxonomy and Future Directions,” IEEE Communications Surveys & Tutorials, vol. 1, 2026.

M. Mamoun and Y. Alhamali, “Deep Reinforcement Learning for Adaptive Traffic Engineering and Congestion Control in Software-Defined Networks,” Texas International Journal of Engineering Research (TIJER), 2026.

A. Muthusamy and R. K. Dhanaraj, “Dynamic Q-Learning-Based Optimized Load Balancing Technique in Cloud,” Mobile Information Systems, vol. 2023, pp. 1–16, Nov. 2023, doi: 10.1155/2023/7250267.

M. M. H. Alkalsh and A. Kliks, “Scalable Mobility Proactive-Reactive Load Balancing (PRLB) Algorithm for High-Density Environments of Future Wireless Networks,” IEEE Access, vol. 13, pp. 140028–140045, 2025, doi: 10.1109/ACCESS.2025.3596682.

K. Chawla, “Reinforcement Learning-Based Adaptive Load Balancing for Dynamic Cloud Environments,” arXiv preprint, Sep. 2024.

V. Mnih et al., “Asynchronous Methods for Deep Reinforcement Learning,” in Proceedings of the 33rd International Conference on Machine Learning (ICML), Feb. 2016.

H. Y. Kim, “An Efficient Load Balancing Scheme for Gaming Server Using Proximal Policy Optimization Algorithm,” Journal of Information Processing Systems, vol. 17, no. 2, pp. 297–305, Apr. 2021, doi: 10.3745/JIPS.03.0158.

I. R. Widiasari and R. Efendi, “Design and Implementation of Load Balancing for Quality of Service Improvement,” Jurnal Buana Informatika, vol. 15, no. 2, pp. 121–132, Oct. 2024.

E. Rahimov and T. Aghayev, “Predictive Load Balancing in Distributed Systems: A Comparative Study of Round Robin, Weighted Round Robin, and a Machine Learning Approach,” in CIEES 2025, Basel Switzerland: MDPI, Jan. 2026, p. 26. doi: 10.3390/engproc2026122026.

Z. Fang, T. Ma, H. Huang, Z. Niu, and F. Yang, “Efficient Task Allocation in Multi-agent Systems by Reinforcement Learning and Genetic Algorithm,” Feb. 04, 2025. doi: 10.20944/preprints202502.0192.v1.

H. Mao, M. Alizadeh, I. Menache, and S. Kandula, “Resource management with deep reinforcement learning,” in HotNets 2016 - Proceedings of the 15th ACM Workshop on Hot Topics in Networks, Association for Computing Machinery, Inc, Nov. 2016, pp. 50–56. doi: 10.1145/3005745.3005750.

J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal Policy Optimization Algorithms,” arXiv preprint, Aug. 2017.

N. Rameshrao Talhar and D. P. Gaikwad, “Adaptive and Optimized Cloud Resource Allocation Using NSGA-II and Proximal Policy Optimization: An Evolutionary Hybrid Genetic-Neural Computing Approach,” Journal of Drug Delivery and Therapeutics, vol. 16, pp. 1053–1068, 2026, doi: 10.25258/ijddt.16.19s.121.

Z. Wu, Z. Jia, X. Pang, and S. Zhao, “Deep Reinforcement Learning-Based Task Offloading and Load Balancing for Vehicular Edge Computing,” Electronics (Switzerland), vol. 13, no. 8, Apr. 2024, doi: 10.3390/electronics13081511.

R. Z. Al-Shaikh, M. M. J. Al-Nayar, and A. M. Hasan, “Reinforcement Learning algorithms for adaptive load-balancing for Web applications,” South African Computer Journal, vol. 37, no. 2, pp. 121–145, 2025, doi: 10.18489/sacj.v37i2.20753.

T. Wang, H. Li, M. Noori, R. Ghiasi, S. C. Kuok, and W. A. Altabey, “Probabilistic Seismic Response Prediction of Three-Dimensional Structures Based on Bayesian Convolutional Neural Network,” Sensors, vol. 22, no. 10, May 2022, doi: 10.3390/s22103775.

M. Noori and W. A. Altabey, “Hysteresis in Engineering Systems,” Oct. 01, 2022, MDPI. doi: 10.3390/app12199428.

R. Prasad and H. Ali, “Dynamic load balancing in distributed systems,” International Journal for Research in Applied Science & Engineering Technology (IJRASET), vol. 13, no. IX, pp. 694–701, Sep. 2025, doi: 10.22214/ijraset.2025.74018.

U. Saha, A. Jawad, S. Shahria, and A. B. M. H. U. Rashid, “Proximal policy optimization-based reinforcement learning approach for DC-DC boost converter control: A comparative evaluation against traditional control techniques,” Heliyon, vol. 10, no. 18, Sep. 2024, doi: 10.1016/j.heliyon.2024.e37823.

E. Altman, CONSTRAINED MARKOV DECISION PROCESSES, 1st Edition. New York: Routledge, 1999.

M. Alshiekh, R. Bloem, R. Udiger Ehlers, B. Könighofer, S. Niekum, and U. Topcu, “Safe Reinforcement Learning via Shielding,” in Proceedings of the 32nd AAAI Conference on Artificial Intelligence, 2018.

C. Zhang, Y. Wang, Y. Lv, H. Wu, and H. Guo, “An Energy and SLA-Aware Resource Management Strategy in Cloud Data Centers,” Scientifics Programming, vol. 2019, 2019, doi: 10.1155/2019/3204346.

M. Chahoud et al., “Reward shaping in DRL: A novel framework for adaptive resource management in dynamic environments,” Informations Sciences, vol. 715, Oct. 2025, doi: 10.1016/j.ins.2025.122238.

C. C. Shetty et al., “CPU-Limits kill Performance: Time to rethink Resource Control,” Association for Computing Machinery (ACM), Nov. 2025, pp. 125–133. doi: 10.1145/3772052.3772219.

D. A. Shafiq, N. Z. Jhanjhi, A. Abdullah, and M. A. Alzain, “A Load Balancing Algorithm for the Data Centres to Optimize Cloud Computing Applications,” IEEE Access, vol. 9, pp. 41731–41744, 2021, doi: 10.1109/ACCESS.2021.3065308.

A. Jitani, A. Mahajan, Z. Zhu, H. Abou-zeid, E. T. Fapi, and H. Purmehdi, “Structure-aware reinforcement learning for node-overload protection in mobile edge computing,” Jun. 2021.

T. Lagkas, P. Sarigiannidis, P. Bellavista, and C. Giannelli, “Multi-stage resource allocation in hybrid 25GEPON and LTE-Advanced Pro FiWi networks for 5G systems,” IET Networks, vol. 7, no. 4, pp. 173–180, Jul. 2018, doi: 10.1049/iet-net.2017.0236.

M. Wang and Z. Gu, “A multi-agent reinforcement learning approach to heterogeneous resource allocation in lifelong education,” International Journal of Information and Communication Technology, vol. 27, no. 7, 2026, doi: 10.1504/IJICT.2026.10075957.

W. Abbass, N. Abbas, and U. Majeed, “SimEdgeAI: A deep reinforcement learning framework for simulating task offloading in resource-constrained IoT networks,” Simul. Model. Pract. Theory, vol. 146, Jan. 2026, doi: 10.1016/j.simpat.2025.103226.

Additional Files

Published

2026-10-10

How to Cite

[1]
A. Ramdan, M. F. Fathoni, and S. Setyorini, “Adaptive Bare-Metal Data Center Load Distribution Using Dual-Policy Deep Reinforcement Learning with Hysteresis Switching and Safety Shielding”, J. Tek. Inform. (JUTIF), vol. 7, no. 5, pp. 4970–4983, Oct. 2026.