Штучний інтелект

Анотація: The rapid and continuous advancement of cloud computing, coupled with increasingly stringent requirements and constraints on cloud services, continues to drive intense research activity in this field. A particularly critical area of study is resource management in cloud environments, which exerts a direct and significant impact on both Quality of Service (QoS) and operational costs. Current mainstream orchestration solutions (such as the Kubernetes Horizontal Pod Autoscaler, HPA) remain predominantly reactive in nature. Consequently, they suffer from resource inefficiencies (over-provisioning and under-provisioning), scaling latencies, and performance degradation. Thus, there is a clear necessity for robust proactive solutions capable of optimizing costs and minimizing scaling delays. This paper provides a comprehensive analysis of state-of-the-art resource orchestration methods in cloud environments and systematizes current approaches to workload forecasting. The efficacy of hybrid models (such as ARIMA-LSTM and Prophet-LSTM) is examined, alongside an analysis of the role of Attention mechanisms and Transformers in the context of multivariate workload prediction. Particular emphasis is placed on the application of Graph Neural Networks (GNNs) as a highly promising direction for modeling complex microservice topologies and predicting latency while accounting for inter-component interactions. Finally, the bottlenecks of existing solutions are identified, and potential vectors for future research are proposed.

Ключові слова: cloud computing, microservices, proactive orchestration, Long Short-Term Memory (LSTM), Transformers, Graph Neural Networks (GNN), workload forecasting, Kubernetes

Посилання:

1. Narváez, D., Battaglia, N., Fernández, A., & Rossi, G. (2025). Designing microservices using AI: A systematic literature review. Software, 4(1), 6. https://doi.org/10.3390/software4010006
2. Nguyen, T.-T., Yeom, Y.-J., Kim, T., et al. (2020). Horizontal pod autoscaling in Kubernetes for elastic container orchestration. Sensors, 20(16), 4621. https://doi.org/10.3390/s20164621
3. Burns, B., Beda, J., Hightower, K., & Evenson, L. (2022). Kubernetes: Up and running: Dive into the future of infrastructure (3rd ed.). O'Reilly Media.
4. Newman, S. (2021). Building microservices: Designing fine-grained systems (2nd ed.). O'Reilly Media.
5. J. Gao, H. Wang and H. Shen, "Machine Learning Based Workload Prediction in Cloud Computing," 2020 29th International Conference on Computer Communications and Networks (ICCCN), Honolulu, HI, USA, 2020, pp. 1-9, doi: 10.1109/ICCCN49398.2020.9209730
6. Arslan S. 2022. A hybrid forecasting model using LSTM and Prophet for energy consumption with decomposition of time series data. PeerJ Computer Science 8: e1001 https://doi.org/10.7717/peerj-cs.1001
7. Zhang, H., Zhang, J., Zeng, J., & Zhu, Z. (2025). Container load prediction method based on a hybrid model of ARIMA and Bi-LSTM. Proc. SPIE 13692, Fourth International Conference on Electronics Technology and Artificial Intelligence (ETAI 2025), 136923M. https://doi.org/10.1117/12.3068497
8. Wang, X., Wu, Y., Zou, W., & Zhao, X. (2025). Hybrid Time Series Forecasting for Real-Time Electricity Market Demand Using ARIMA-LSTM and Scalable Cloud-Native Architecture. Informatica, 49(3). https://doi.org/10.31449/inf.v49i3.9474
9. Klymash, M., & Morhoiev, K. (2025). Method for workload-based selection of lightweight prediction models in microservice autoscaling. ICTEE, 5(2), 1–14. https://doi.org/10.23939/ictee2025.02.001
10. Ahamed, Z., et al. (2023). Technical study of deep learning in cloud computing for accurate workload prediction. Electronics, 12(3), 650. https://doi.org/10.3390/electronics12030650
11. Vasumathi, M. T., et al. (2025). AI-driven predictive auto-scaling for efficient microservices deployment in cloud data centers using LSTM and Kubernetes HPA. 2025 5th International Conference on Expert Clouds and Applications (ICOECA), 859–864. https://doi.org/10.1109/ICOECA66273.2025.00151
12. Patel, Y. S., & Bedi, J. (2023). MAG-D: A multivariate attention network based approach for cloud workload forecasting. Future Generation Computer Systems, 142, 376–392. https://doi.org/10.1016/j.future.2023.01.002
13. Sabyasachi, A. S., Sahoo, B. M., & Ranganath, A. (2024). Deep CNN and LSTM approaches for efficient workload prediction in cloud environment. Procedia Computer Science, 235, 2651–2661. https://doi.org/10.1016/j.procs.2024.04.250
14. Zhang, H., Li, J., & Yang, H. (2024). Cloud computing load prediction method based on CNN-BiLSTM model under low-carbon background. Scientific Reports, 14, 18004. https://doi.org/10.1038/s41598-024-68339-1
15. Mahajan, I., & Nadig, D. (2024). Enhancing workload predictions using service interactions in cloud-native microservices. 2024 IEEE 13th International Conference on Cloud Networking (CloudNet), 1–9. https://doi.org/10.1109/CloudNet62863.2024.10815917
16. Chauhan, R., Yang, J., & Khan, A. A. (2026). Reducing the carbon footprint of data centers: CNN-transformer-based workload prediction for green cloud computing. In Computational Intelligence and Network Security. ICCINS 2025 (Vol. 2738, pp. 222–235). Springer. https://doi.org/10.1007/978-3-032-09572-5_22
17. Kumar, B., Verma, A., & Verma, P. (2025). A multivariate transformer-based monitor-analyze-plan-execute (MAPE) autoscaling framework for dynamic resource allocation in cloud environment. Computing, 107, 69. https://doi.org/10.1007/s00607-025-01426-x
18. Park, J., et al. (2024). Graph neural network-based SLO-aware proactive resource autoscaling framework for microservices. IEEE/ACM Transactions on Networking, 32(4), 3331–3346. https://doi.org/10.1109/TNET.2024.3393427
19. Luo, Y., et al. (2024). Integrating system state into spatio temporal graph neural network for microservice workload prediction. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '24), 5521–5531. https://doi.org/10.1145/3637528.3671508
20. Li, J., et al. (2024). EvoGWP: Predicting long-term changes in cloud workloads using deep graph-evolution learning. IEEE Transactions on Parallel and Distributed Systems, 35(3), 499–516. https://doi.org/10.1109/TPDS.2024.3357715
21. Tam, D. S. H., et al. (2023). PERT-GNN: Latency prediction for microservice-based cloud-native applications via graph neural networks. Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '23), 2155–2165. https://doi.org/10.1145/3580305.3599465
22. Tam, D. S. H., et al. (2025). FastPERT: Towards fast microservice application latency prediction via structural inductive bias over PERT networks. Proceedings of the AAAI Conference on Artificial Intelligence, 39(19), 20787–20795. https://doi.org/10.1609/aaai.v39i19.34291