绿色计算下算力调度优化问题与技术研究

doi:10.15960/j.cnki.issn.1007-6093.2025.03.009

摘要/Abstract

摘要： 数字经济时代,随着云计算与人工智能行业的飞速发展,算力作为重要战略资源,价值日益凸显,算力应用所产生的能耗和碳排放量也在急剧攀升。在此背景下,绿色计算的发展已成为行业共识和时代需求,算力资源的调度优化也成为节能减排、降本增效的重要手段。本文重点研究了绿色计算应用场景中的4类具体的算力调度优化问题:计算任务错峰调度、容器负载均衡、集群自动扩缩容、服务混合均匀部署,给出了这几类调度优化问题对应的数学模型和优化算法,并进一步介绍了工业场景下的智能算力调度系统和落地挑战。这套算力调度系统已经服务于蚂蚁集团大数据计算、数据库等多个应用场景,为企业节能减排带来了显著收益。最后,本文展望了算力调度在AI大模型时代下的挑战。

关键词: 绿色计算, 算力调度, 负载均衡, 数据中心, 人工智能

Abstract: In the era of the digital economy, the rapid development of cloud computing and artificial intelligence industries has made computing power an increasingly valuable strategic resource. However, the energy consumption and carbon emissions generated by computing applications are also rising sharply. In this context, the development of green computing has become an industry consensus and a necessity of our times. Optimizing the scheduling of computing resources has emerged as a critical approach to reducing energy consumption, lowering costs, and improving efficiency. This paper focuses on four specific optimization problems related to computing resource scheduling: peak-shifting scheduling of computational tasks, load balancing of containers, autoscaling of clusters, and uniform deployment of mixed services. For each of these optimization problems, we present corresponding mathematical models and optimization algorithms. Furthermore, we introduce an intelligent computing power scheduling system designed for industrial applications, along with the challenges faced during its implementation. This scheduling system has been successfully applied to various scenarios within Ant Group, including big data computing and database management, delivering significant benefits in terms of carbon neutrality and energy savings. The results demonstrate that the system not only reduces energy consumption but also enhances operational efficiency, offering a practical solution for enterprises to achieve sustainability goals. Finally, this paper discusses the challenges of computing power scheduling in the era of large AI models, where the growing complexity and scale of computations demand even more innovative approaches.

Key words: green computing, computing power scheduling, load balancing, data center, artificial intelligence

中图分类号:

O221.7

鲁炜, 卢星宇, 邹丁, 陈博晓, 周义涵, 张国川. 绿色计算下算力调度优化问题与技术研究[J]. 运筹学学报（中英文）, 2025, 29(3): 179-201.

LU Wei, LU Xingyu, ZOU Ding, CHEN Boxiao, ZHOU Yihan, ZHANG Guochuan. Research on computing power scheduling problem and technology in green computing[J]. Operations Research Transactions, 2025, 29(3): 179-201.

参考文献

[1] Marzouk O A. Summary of the 2023 report of TCEP (tracking clean energy progress) by the international energy agency (IEA), and proposed process for computing a single aggregate rating [C]//E3S Web of Conferences, 2025, 601: 00048.
[2] 中国信息通信研究院. 中国绿色算力发展研究报告(2024年)[R]. 北京: 中国信息通信研究院, 2024.
[3] 蚂蚁集团. 蚂蚁集团2021年可持续发展报告[R]. 杭州: 蚂蚁集团, 2022.
[4] 蚂蚁集团, 中国信通院. 面向算力应用环节的计算绿色化白皮书[R]. 杭州: 蚂蚁集团, 2023.
[5] 蚂蚁集团. 蚂蚁集团2023年可持续发展报告[R]. 杭州: 蚂蚁集团, 2024.
[6] Arshad M, Hellerstein J L, Evers C T. A survey of auto-scaling techniques for elastic applications in cloud computing [J]. IEEE Transactions on Network and Service Management, 2018, 15(1): 36-52.
[7] Khan S A, Ahmad I, Madria S. Resource provisioning in cloud computing environments: A survey [J]. ACM Computing Surveys, 2014, 46(4): 1-36.
[8] Aslam A M, Jaur M. A review on energy efficient technique in green cloud: Open research challenges and issues [J]. International Journal on Scientific Research in Computer Science and Engineering, 2018, 6(3): 44-50.
[9] Yang Z, Qin X, Li W, et al. Optimized task scheduling and resource allocation in cloud computing using PSO based fitness function [J]. Information Technology Journal, 2013, 12(23): 7090.
[10] Zhao G. Cost-aware scheduling algorithm based on PSO in cloud computing environment [J]. International Journal of Grid & Distributed Computing, 2014, 7: 33-42.
[11] Ghamkhari M, Mohsenian-Rad H. Optimal integration of renewable energy resources in data centers with behind-the-meter renewable generator [C]//Proceedings of the 2012 IEEE International Conference on Communications, 2012.
[12] Beloglazov A, Buyya R. Energy efficient allocation of virtual machines in cloud data centers [C]//Proceedings of the 201010th IEEE/ACM International Conference on Cluster, Cloud and Grid Computing, 2010: 577-578.
[13] Huang J, Wu K, Moh M. Dynamic virtual machine migration algorithms using enhanced energy consumption model for green cloud data centers [C]//Proceedings of the 2014 International Conference on High Performance, 2014.
[14] Dupont C, Schulze T, Giuliani G, et al. An energy aware framework for virtual machine placement in cloud federated data centres [C]//Proceedings of the 3rd International Conference on Future Energy Systems: Where Energy, Computing and Communication Meet, 2012: 1-10.
[15] Copil G, Moldovan D, Salomie I, et al. Cloud SLA negotiation for energy saving|A particle swarm optimization approach [C]//Proceedings of the 2012 IEEE 8th International Conference on Intelligent Computer Communication and Processing, 2012.
[16] Kavitha K, Kumuthini C. Energy efficient computing strategies for green cloud [J]. Tuijin Jishu/Journal of Propulsion Technology, 2023, 44(6): 6009-6016.
[17] Valancius V, Laoutaris N, Massoulié L, et al. Greening the internet with NanoData Centers [C]//Proceedings of the 2009 CoNEXT Conference, 2009: 1-12.
[18] Chen G, He W, Liu J, et al. Energy-aware server provisioning and load dispatching for connection-intensive internet services [C]//Proceedings of the 5th USENIX Symposium on Networked Systems Design and Implementation, 2008: 337-350.
[19] Shu W, Wang W, Wang Y. A novel energy-efficient resource allocation algorithm based on immune clonal optimization for green cloud computing [J]. EURASIP Journal on Wireless Communications and Networking, 2014, 2014: 64.
[20] Liu Y, Shu W, Zhang C. A parallel task scheduling optimization algorithm based on clonal operator in green cloud computing [J]. Journal of Communications, 2016, 11: 185-191.
[21] Dhingra A. Green cloud: Smart resource allocation and optimization using simulated annealing technique [J]. Indian Journal of Computer Science and Engineering, 2014, 5(2): 1-10.
[22] Okewu E, Misra S, Maskelijnas R, et al. Optimizing green computing awareness for environmental sustainability and economic security as a stochastic optimization problem [J]. Sustainability, 2017, 9(10): 1857.
[23] Garg S K, Yeo C S, Anandasivam A, et al. Environment-conscious scheduling of HPC applications on distributed cloud-oriented data centers [J]. Journal of Parallel and Distributed Computing, 2011, 71: 732-749.
[24] Zhang L M, Li K, Zhang Y Q. Green task scheduling algorithms with speeds optimization on heterogeneous cloud servers [C]//Proceedings of the 2010 IEEE/ACM Int’l Conference on Green Computing and Communications & Int’I Conference on Cyber, Physical and Social Computing, 2010: 76-80.
[25] Borgetto D, Casanova H, Da Costa G, et al. Energy-aware service allocation [J]. Future Generation Computer Systems, 2012, 28(5): 769-779.
[26] Cabrera G, Pérez-Rosés H, Juan A A, et al. Operations research in green internet computing: State of the art and open challenges [J]. International Journal of Production Research, 2013.
[27] Zou D, Lu W, Zhu Z, et al. OptScaler: A collaborative framework for robust autoscaling in the cloud [J]. Proceedings of the VLDB Endowment, 2024, 17(12): 4090-4103.
[28] Dong H, Wang B, Qiao B, et al. Predictive job scheduling under uncertain constraints in cloud computing [C]//Proceedings of the 30th International Joint Conference on Artificial Intelligence, 2021: 3627-3634.
[29] Qiao B, Yang F, Luo C, et al. Intelligent container reallocation at Microsoft 365[C]//Proceedings of the 29th ACM Joint Meeting on European Software Engineering Conference and Symposium on the Foundations of Software Engineering, 2021: 1438-1443.
[30] Lu X, Liu Z, Guan Y, et al. GreenFlow: A computation allocation framework for building environmentally sound recommendation system [C]//Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023: 6103-6111.
[31] Rzadca K, Findeisen P, Swiderski J, et al. Autopilot: Workload autoscaling at Google [C]//Proceedings of the Fifteenth European Conference on Computer Systems, 2020: 1-16.
[32] Zou D, Lian J, Lu W, et al. An equally-split bin packing problem [C]//Proceedings of International Conference on Combinatorial Optimization and Applications, 2025: 240-252.
[33] Hart P E, Nilsson N J, Raphael B. A formal basis for the heuristic determination of minimum cost paths [J]. IEEE Transactions on Systems Science and Cybernetics, 1968, 4(2): 100-107.
[34] Christophere B. Model predictive control: Principles and case studies [M]//Model Predictive Control: Theory and Application, Oxford: Oxford University Press, 1995.
[35] Shahrad M, Fonseca R, Goiri I, et al. Serverless in the wild: characterizing and optimizing the serverless workload at a large cloud provider [C]//Proceedings of the 2020 USENIX Annual Technical Conference, 2020: 205-218.
[36] JV B B, Dharma D. HAS: Hybrid auto-scaler for resource scaling in cloud environment [J]. Journal of Parallel and Distributed Computing, 2018, 120: 1-15.
[37] Tianchi. Alibaba data center resource management and scheduling dataset [EB/OL]. [2025- 03-01]. https://tianchi.aliyun.com/dataset/dataDetail?datald=6287.
[38] Mommessin C, Yang R, Shakhlevich N V, et al. Affinity-aware resource provisioning for longrunning applications in shared clusters [J]. Journal of Parallel and Distributed Computing, 2023, 177: 1-16.
[39] Cole D. Data center energy efficiency|looking beyond PUE [R]. No Limits Software, White Paper, 2011.
[40] Delage E, Ye Y. Distributionally robust optimization under moment uncertainty with application to data-driven problems [J]. Operations Research, 2010, 58(3): 595-612.
[41] Chen M, Xu Z, Wang Y, et al. Deep reinforcement learning for large-scale logistics scheduling [C]//Proceedings of the 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2020: 3452-3461.
[42] Zhang Y, Liu Q, Wu J, et al. Multi-agent deep reinforcement learning for power grid dispatch with high renewable penetration [J]. IEEE Transactions on Sustainable Energy, 2022, 13(3): 1457-1469.
[43] Tang H, Liu J, Zhou Z, et al. Graph reinforcement learning for ride-hailing order dispatching [C]//Proceedings of the 35th AAAI Conference on Artificial Intelligence, 2021: 12361-12369.