Yaqi Hou
School of Information Science and Engineering, Central South University, Changsha 410083, Hunan, China.
*Corresponding author: Yaqi Hou
Abstract
In the operating context where multi-generation CPUs and GPUs are deployed in parallel for a long time, the cost structure of data centers has gradually shifted from a single hardware investment issue to a comprehensive economic problem intertwined with resource scheduling, energy efficiency control, and lifecycle management. The differences in computing power density, power consumption curves, and task types among heterogeneous chips make the traditional configuration approach centered on hardware performance difficult to achieve stable cost control targets. Based on the evolution characteristics of computing power deployment and operational data features, this paper starts from the relationship between the cost composition mechanism and scheduling behavior, and constructs an optimization path framework centered on the task mapping mechanism, energy efficiency evaluation rules, and the cost feedback loop. It reveals the intrinsic relationship between resource allocation and economy in multi-generation collaborative architectures. This path emphasizes dynamic adjust-ment based on the operating status, making the computing power deployment strategy have both performance orientation and cost constraint attributes, providing a sustainable economic management idea for data center operations in high-density computing environments.
References
[1] Chatzistefanidis I, Nikaein N. Symbiotic agents: A novel paradigm for trustworthy AGI-driven networks. Comput Netw. 2025;273:111749.
[2] Song X. Research on optimization of e-commerce supply chain logistics management model based on blockchain. Front Soc Sci Technol. 2025;7(3).
[3] Dong Z, Ge X, Huang Y, et al. EG-STC: An Efficient Secure Two-Party Computation Scheme Based on Embedded GPU for Artificial Intelligence Systems. Comput Mater Contin. 2024;79(3):4021-44.
[4] Nuno F. GPU acceleration of the ATLAS calorimeter clustering algorithm. J Phys Conf Ser. 2023;2438(1):012023.
[5] Chiheb AA, Wael L, Faisal M, et al. Firefly algorithm and learning-based geographical task scheduling for operational cost mini-mization in distributed green data centers. Neurocomputing. 2022;490:146-62.
[6] H R, P S. Analytical Comparison of Cloud Data Centre Services and Cost. IOP Conf Ser Mater Sci Eng. 2021;1022(1):012058.
[7] Kumar M, Sharma CS, Goel S, et al. Autonomic cloud resource provisioning and scheduling using meta-heuristic algorithm. Neural Comput Appl. 2020;32(24):18253-71.
[8] Tang W, Cai W, Yao Y, et al. An alternative approach for collaborative simulation execution on a CPU+GPU hybrid system. Simulation. 2020;96(3):347-61.
[9] Yeganeh H, Salahi A, Pourmina AM. A Novel Cost Optimization Method for Mobile Cloud Computing by Capacity Planning of Green Data Center With Dynamic Pricing. Can J Electr Comput Eng. 2019;42(1):41-51.
How to cite this paper
Exploration of Data Center Cost Optimization Pathways Under Multi-generation CPU and GPU Collaborative Architectures
How to cite this paper: Yaqi Hou. (2026). Exploration of Data Center Cost Optimization Pathways Under Multi-generation CPU and GPU Collaborative Architectures. Engineering Advances, 6(1), 41-44.
DOI: http://dx.doi.org/10.26855/ea.2026.03.009