With the widespread application of cloud desktop systems, virtual machine scheduling algorithms face the challenge of effectively handling complex and dynamic workloads. Traditional scheduling algorithms often underperform in terms of resource utilization, system latency, and load balancing. This paper proposes a cloud desktop virtual machine scheduling optimization algorithm that combines graph neural networks (GNN) and reinforcement learning (RL). We model the virtual machines and their resource requirements in the cloud desktop environment as a graph structure, and use GNN to predict the load conditions of the virtual machines. By incorporating RL strategies, we dynamically adjust resource allocation and virtual machine migration decisions based on the prediction results to optimize system performance. The algorithm is evaluated on multiple datasets from real-world scenarios, including 4K video processing, office applications, and network applications, by measuring indicators like resource utilization, system latency, and load balancing. Experimental results show that the proposed scheduling algorithm exhibits significant improvements across multiple datasets. Compared to traditional algorithms, resource utilization increases by over 12%, system latency is reduced by 15%, and load balancing is significantly better.
对云计算环境中动态任务调度的鲁棒性和截止时间保障问题,文献[19]提出了一种基于元深度强化学习(Meta Deep Reinforcement Learning, MDRL)的调度解决方案。该方法通过量化鲁棒性指标(如重新训练时间),在高度动态的任务负载和资源可用性变化下,提升了调度性能的鲁棒性和适应速度。
为全面评估算法性能,我们实现了如下算法的对比:传统负载均衡算法,包括轮询算法(Round Robin,RR)、最短作业优先算法(Shortest Job First,SJF)和先来先服务算法(First Come First Served,FCFS);基于GA启发式虚拟机调度,以及基于机器学习随机森林(Random Forest, RF)的虚拟机调度方法。
FANGY Q, TANGD H, GEJ W. Research on Schedule Strategy Based on Dynamic Migration of Virtual Machines in Cloud Environment[J]. Microelectron Comput, 2012, 29(4): 45-48. DOI: 10.19304/j.cnki.issn1000-7180.2012.04.011 .
SINGHG, KAURK. An Improved Weighted Least Connection Scheduling Algorithm for Load Balancing in Web Cluster Systems[J]. Int Res J Eng Technol, 2018, 5(3): 6.
ZHANGY J, HEF, LUG S, et al. A Modified Weighted round Robin Scheduling Algorithm in TTE[J]. J Beijing Univ Aeronaut Astronaut, 2017, 43(8): 1577-1584. DOI: 10.13700/j.bh.1001-5965.2016.0590 .
[8]
XUH, XUS, WEIW, et al. Fault Tolerance and Quality of Service Aware Virtual Machine Scheduling Algorithm in Cloud Data Centers[J]. J Supercomput, 2023, 79(3): 2603-2625. DOI:10.1007/s11227-022-04760-5 ..
ZHANGZ Y, ZHOUR L, SUNB, et al. Load Balancing Algorithm for Data Distribution Network Based on XMPP Protocol[J]. J Beijing Univ Posts Telecommun, 2016, 39(S1): 27-31. DOI: 10.13190/j.jbupt.2016.s.007 .
YANX Y, HANP. Research and Application of Load Balance Technology in Cloud Computing Environment[J]. Microelectron Comput, 2015, 32(10): 181-184. DOI: 10.19304/j.cnki.issn1000-7180.2015.10.039 .
[13]
CHITRA DEVID, RHYMEND UTHARIARAJV. Load Balancing in Cloud Computing Environment Using Improved Weighted round Robin Algorithm for Nonpreemptive Dependent Tasks[J]. Sci World J, 2016, 2016: 3896065. DOI:10.1155/2016/3896065 .
XUS C. Greedy Algorithms Optimized Virtual Machine Allocation for Cloud Data Centers[J]. Comput Syst Appl, 2021, 30(3): 134-141. DOI: 10.15888/j.cnki.csa.007814 .
[16]
WUG, TANGM L, TIANY C, et al. Energy-efficient Virtual Machine Placement in Data Centers by Genetic Algorithm[M]//Neural Information Processing. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012: 315-323. DOI:10.1007/978-3-642-34487-9_39 .
[17]
TSENGF H, CHENX J, CHOUL D, et al. Support Vector Machine Approach for Virtual Machine Migration in Cloud Data Center[J]. Multimed Tools Appl, 2015, 74(10): 3419-3440. DOI:10.1007/s11042-014-2086-z .
[18]
KUMARK V, MALATHIP, RAMESHS. Performance Analysis of Placement Prediction System Using Support Vector Machine over Random Forest Algorithm[C]//2022 14th International Conference on Mathematics, Actuarial Science, Computer Science and Statistics (MACS). New York: IEEE, 2022: 1-4. DOI:10.1109/MACS56771.2022.10023271 .
[19]
RAMANN, WAHABA B, CHANDRASEKARANS. Computation of Workflow Scheduling Using Backpropagation Neural Network in Cloud Computing: A Virtual Machine Placement Approach[J]. J Supercomput, 2021, 77(9): 9454-9473. DOI:10.1007/s11227-021-03648-0 .
SUNP H, LANJ L, LIJ F, et al. Combining Deep Reinforcement Learning with Graph Neural Networks for Optimal VNF Placement[J]. IEEE Commun Lett, 2020, 25(1): 176-180. DOI:10.1109/LCOMM.2020.3025298 .
[22]
SCARSELLIF, GORIM, TSOIA C, et al. The Graph Neural Network Model[J]. IEEE Trans Neural Netw, 2009, 20(1): 61-80. DOI:10.1109/TNN.2008.2005605 .
[23]
ZHANGY F, LIUB, GONGY L, et al. Application of Machine Learning Optimization in Cloud Computing Resource Scheduling and Management[C]//Proceedings of the 5th International Conference on Computer Information and Big Data Application. New York: Association for Computing Machinery, 2024.
[24]
LIUH Y, CHENP, OUYANGX, et al. Robustness Challenges in Reinforcement Learning Based Time-critical Cloud Resource Scheduling: A Meta-learning Based Solution[J]. Future Gener Comput Syst, 2023, 146: 18-33. DOI:10.1016/j.future.2023.03.029 .
DANGW C, WUT Y. Active Fault-tolerant Priority Migration Decision Model for Virtual Machines Based on Attentional Spatio-temporal Convolution and A2C[J]. Appl Res Comput, 2023, 40(12): 3606-3613. DOI: 10.19734/j.issn.1001-3695.2023.03.0144 .
[27]
KALASKARC, THANGAMS. A Graph Neural Network-based Approach with Dynamic Multiqueue Optimization Scheduling (DMQOS) for Efficient Fault Tolerance and Load Balancing in Cloud Computing[J]. Int J Intell Syst, 2024, 2024(1): 6378720. DOI:10.1155/int/6378720 .
[28]
SIMAIYAS, LILHOREU K, SHARMAY K, et al. A Hybrid Cloud Load Balancing and Host Utilization Prediction Method Using Deep Learning and Optimization Techniques[J]. Sci Rep, 2024, 14(1): 1337. DOI:10.1038/s41598-024-51466-0 .
[29]
ZHANGZ, XUC, LIUK, et al. A Resource Optimization Scheduling Model and Algorithm for Heterogeneous Computing Clusters Based on GNN and RL[J]. J Supercomput, 2024, 80(16): 24138-24172. DOI:10.1007/s11227-024-06383-4 .
[30]
MENGQ F, LUOZ H, ZHENGX L, et al. Hierarchical Collaborative Resource Scheduling in Industrial Internet of Things Based on Graph Neural Networks and Deep Reinforcement Learning[C]//2023 IEEE 29th International Conference on Parallel and Distributed Systems (ICPADS). New York: IEEE, 2023: 2647-2654. DOI:10.1109/ICPADS60453.2023.00351 .
[31]
KINGC I. Stress-ng (Version 0.17.09)[CP/OL]. (2024-04-16)[2024-12-01].