基于LSTM和遗传算法的容器资源智能调度策略研究
Research on Intelligent Scheduling Strategy of Container Resources Based on LSTM and Genetic Algorithm
DOI: 10.12677/csa.2024.1412247, PDF,   
作者: 梁 拓, 王利众：中央民族大学信息工程学院，北京；许 震：北京青云科技集团股份有限公司项目服务部门，北京
关键词: 容器资源调度；LSTM；遗传算法；Kubernetes；自动伸缩；Container Resource Scheduling； LSTM； Genetic Algorithm； Kubernetes； Automatic Scaling
摘要: 本研究旨在优化Kubernetes集群中容器化服务的资源调度，提高资源利用效率和系统性能。针对现有Kubernetes自动伸缩机制在应对复杂负载时的不足，提出一种基于LSTM和遗传算法的智能资源调度策略。首先利用LSTM模型预测未来的资源需求，以应对不同的负载波动；然后利用遗传算法优化资源配额，确保在高效利用资源的同时满足系统性能要求。实验在实际Kubernetes集群环境中进行，设置高负载、突发请求和低负载等负载场景，对比LSTM+遗传算法与Kubernetes默认HPA/VPA方案在响应时间、吞吐量和资源利用率方面的性能。结果表明，LSTM+遗传算法在大多数负载场景下均优于HPA/VPA策略，提高了平均资源利用率，降低了系统响应时间。
Abstract: This study aims to optimize the resource scheduling of containerized services in Kubernetes clusters and improve resource utilization efficiency and system performance. In view of the shortcomings of the existing Kubernetes auto-scaling mechanism in dealing with complex loads, an intelligent resource scheduling strategy based on LSTM and genetic algorithm is proposed. First, the LSTM model is used to predict future resource requirements to cope with different load fluctuations; then the genetic algorithm is used to optimize resource quotas to ensure that system performance requirements are met while efficiently utilizing resources. The experiment is carried out in an actual Kubernetes cluster environment, setting load scenarios such as high load, burst requests, and low load, and comparing the performance of LSTM+ genetic algorithm and Kubernetes default HPA/VPA solution in terms of response time, throughput, and resource utilization. The results show that LSTM+ genetic algorithm outperforms HPA/VPA strategy in most load scenarios, improves average resource utilization, and reduces system response time.
文章引用：梁拓, 王利众, 许震. 基于LSTM和遗传算法的容器资源智能调度策略研究[J]. 计算机科学与应用, 2024, 14(12): 132-141. https://doi.org/10.12677/csa.2024.1412247

参考文献

[1] Burns, B., Grant, B., Oppenheimer, D., Brewer, E. and Wilkes, J. (2016) Borg, Omega, and Kubernetes. Communications of the ACM, 59, 50-57. [Google Scholar] [CrossRef] 
[2] Xu, Z., Gong, Y., Zhou, Y., Bao, Q. and Qian, W. (2024). Enhancing Kubernetes Automated Scheduling with Deep Learning and Reinforcement Techniques for Large-Scale Cloud Computing Optimization. 9th International Symposium on Advances in Electrical, Electronics, and Computer Engineering (ISAEECE 2024), Volume 13291, 132916E.[CrossRef] 
[3] Nguyen, T., Yeom, Y., Kim, T., Park, D. and Kim, S. (2020) Horizontal Pod Autoscaling in Kubernetes for Elastic Container Orchestration. Sensors, 20, Article No. 4621. [Google Scholar] [CrossRef] [PubMed]
[4] Niazi, M., Abbas, S., Soliman, A., Alyas, T., Asif, S. and Faiz, T. (2023) Vertical Pod Autoscaling in Kubernetes for Elastic Container Collaborative Framework. Computers, Materials & Continua, 74, 591-606. [Google Scholar] [CrossRef] 
[5] Bi, J., Li, S., Yuan, H. and Zhou, M. (2021) Integrated Deep Learning Method for Workload and Resource Prediction in Cloud Systems. Neurocomputing, 424, 35-48. [Google Scholar] [CrossRef] 
[6] Huang, C., Wang, Y., Guan, C., Chen, H. and Jian, J. (2013) Applications of Machine Learning to Resource Management in Cloud Computing. International Journal of Modeling and Optimization, 2, 148-152. [Google Scholar] [CrossRef] 
[7] Graves, A. (2012) Long Short-Term Memory. In: Graves, A., Ed., Supervised Sequence Labelling with Recurrent Neural Networks, Springer, 37-45. [Google Scholar] [CrossRef] 
[8] Lindemann, B., Müller, T., Vietz, H., Jazdi, N. and Weyrich, M. (2021) A Survey on Long Short-Term Memory Networks for Time Series Prediction. Procedia CIRP, 99, 650-655. [Google Scholar] [CrossRef] 
[9] Thonglek, K., Ichikawa, K., Takahashi, K., Iida, H. and Nakasan, C. (2019). Improving Resource Utilization in Data Centers Using an LSTM-Based Prediction Model. 2019 IEEE International Conference on Cluster Computing (CLUSTER), Albuquerque, 23-26 September 2019, 1-8.[CrossRef] 
[10] Holland, J.H. (1975) Adaptation in Natural and Artificial Systems. University of Michigan Press.
[11] Gawanmeh, A., Parvin, S. and Alwadi, A. (2017) A Genetic Algorithmic Method for Scheduling Optimization in Cloud Computing Services. Arabian Journal for Science and Engineering, 43, 6709-6718. [Google Scholar] [CrossRef]