A Hybrid Machine Learning And Reinforcement Learning Framework For Energy-Aware Resource Scheduling With Pue Integration
Main Article Content
Abstract
Energy efficiency in data centers is predominantly evaluated through Power Usage Effectiveness (PUE), yet existing approaches often neglect the complex interplay between workload dynamics, thermal profiles, and hardware reliability. This gap results in suboptimal optimization strategies that focus narrowly on energy ratios rather than system-wide cost efficiency. To address this limitation, we present a machine learning–driven resource scheduling framework that bridges PUE optimization with holistic system modeling. The framework integrates workload forecasting via gradient boosted decision trees, thermal prediction through long short-term memory networks, and reinforcement learning–based control for real-time, multi-objective optimization of compute allocation and cooling strategies. Validation using a DCIM-based digital twin, real workload traces, and ASHRAE- compliant cooling profiles demonstrates up to 17.4% reduction in total energy cost, 12.6% decrease in hardware degradation cost, and 15.2% improvement in cooling efficiency without violating service-level agreements. Owing to its modular and hardware-agnostic design, the framework ensures scalability and practical deployment in production-scale data centers.
Article Details
References
Z. Li, H. Xu, and Y. Wang, “AI-based predictive scheduling for large-scale cloud data centers,” IEEE Transactions on Cloud Computing, vol. 11, no. 1, pp. 85–97, Jan. 2023.
S. Ahmed and R. Gupta, “Adaptive workload-aware energy optimization in edge-cloud environments,”Future Generation Computer Systems, vol. 138, pp. 75–88, Apr. 2023.
M. Chen, X. Li, and C. Yang, “Deep reinforcement learning for resource allocation in mobile edgecomputing,” IEEE Internet of Things Journal, vol. 10, no. 4, pp. 3201–3215, Feb. 2023.
R. Singh and P. Kumar, “QoS-driven resource orchestration in distributed computing platforms,” Journal of Parallel and Distributed Computing, vol. 176, pp. 42–55, May 2023.
L. Zhao, J. Zhang, and K. Yang, “Carbon-aware workload scheduling for sustainable data centers,” IEEE Transactions on Sustainable Computing, vol. 8, no. 2, pp. 156–169, Mar. 2023.
T. Nguyen, H. Le, and M. Tran, “Hybrid optimization for energy-efficient resource scheduling in fogcomputing,” Journal of Network and Computer Applications, vol. 214, 103515, Jul. 2023.
D. Patel, S. Bose, and V. Sharma, “Federated learning-enabled resource management for energy-aware edge systems,” IEEE Transactions on Green Communications and Networking, vol. 7, no. 3, pp. 1451–1463, Sep. 2023.
M. Al-Khafaji and Y. Wang, “Task offloading and resource allocation using evolutionary algorithms in IoT- enabled edge systems,” Applied Soft Computing, vol. 144, 110342, Oct. 2023.
C. Li, J. Yu, and A. Ghosh, “Holistic system modeling for multi-resource optimization in heterogeneousclouds,” IEEE Transactions on Services Computing, vol. 17, no. 1, pp. 30–44, Jan. 2024.
S. Roy, F. Alam, and R. J. Figueiredo, “Renewable energy-aware cloud resource management: Trends and future challenges,” Sustainable Energy Technologies and Assessments, vol. 60, 103563, Feb. 2024.
B. Kumar and A. Verma, “Multi-objective ML-driven scheduling for latency and energy optimization,”Future Generation Computer Systems, vol. 149, pp. 224–239, Mar. 2024.
L. Tang, M. Huang, and Z. Chen, “Integrating edge and renewable energy for sustainable IoT ecosystems,”IEEE Transactions on Industrial Informatics, vol. 20, no. 5, pp. 6011–6022, May 2024.
R. Zhang, H. Wu, and S. Li, “Energy-efficient resource provisioning with carbon-intensity modeling,”Journal of Cleaner Production, vol. 425, 139821, Jun. 2024.
Y. Han and K. Lee, “AI-augmented holistic modeling for large-scale distributed systems,” IEEE Transactions on Parallel and Distributed Systems, vol. 35, no. 7, pp. 921–934, Jul. 2024.
G. Zhao and W. Chen, “Carbon-neutral edge-cloud collaboration for green computing,” IEEE Transactions on Green Communications and Networking, vol. 8, no. 4, pp. 2335–2348, Aug. 2024.
A. Sharma and M. Patel, “PUE-aware optimization in hybrid renewable-powered data centers,” Energy Reports, vol. 11, pp. 3402–3416, Aug. 2024.
S. Banerjee, X. Liu, and R. Khan, “ML-enabled adaptive cooling strategies for sustainable data centers,”IEEE Access, vol. 12, pp. 113021–113035, Sep. 2024.
T. Wang and D. Xu, “Holistic scheduling framework for workload and renewable integration in edge-cloud systems,” Future Generation Computer Systems, vol. 154, pp. 187–201, Oct. 2024.
P. Nair, V. Gupta, and L. Singh, “Reinforcement learning-based multi-tier scheduling with PUEconstraints,” IEEE Transactions on Cloud Computing, Early Access, Nov. 2024.
Y. Zhang, J. Chen, and L. Wu, “AI-driven holistic energy-aware scheduling: Bridging theory and practice,”IEEE Transactions on Sustainable Computing, Early Access, Dec. 2024.
Dewangan, O., & Sarkar, P. (2022, December), “A Study on Network Security Using Deep Learning Methods,” Advanced Engineering Science, 54(02), 6393 - 6404.
Sarkar, P., & Dewangan, O. (2023), “Augmented reality-based virtual smartphone,” Journal of Data Acquisition and Processing, 38(2), 1983-1990.
Shanmugavelu, A. K. T., Muraliraja, R., Shanmugam, R., Pawar, M. P. S., Vishwakarma, R., & Sarkar, P. (2023), “Design of Subsea storage tanks for Arctic conditions - heat treatment of materials,” Materials Today: Proceedings. https://doi.org/10.1016/j.matpr.2023.04.367.
Sarkar, P. (2023), “The Future is Now: Exploring the Role of AI in Biochemical Structure Analysis,” European Chemical Bulletin, vol. 12 (Special Issue 1), pp. 5104-5116.
H. Kim and J. Park, “Towards zero-carbon edge-cloud systems with intelligent workload distribution,”Applied Energy, vol. 356, 122483, Dec. 2024.