This article presents a systematic analysis of architectural patterns for integrating AI-driven optimization into high-load distributed systems operating in production environments. Such systems, built on cloud-native, containerized, and edge platforms, are characterized by dynamic workloads, strict service-level requirements, and high sensitivity to control errors, which limit the effectiveness of reactive and rule-based orchestration mechanisms. The study is conducted as a review-and-analytical synthesis of peer-reviewed publications, focusing on architectural placement of intelligent components, types of control loops, and operational constraints, without quantitative aggregation of results due to heterogeneity of experimental settings and metrics. Particular attention is paid to production-oriented optimization patterns that preserve standard orchestration mechanisms while constraining or parameterizing their decision space, as well as to approaches that introduce intelligence into scheduling and workflow-level coordination. The analysis highlights the trade-offs between measurable performance and cost gains, increased architectural complexity, control-loop stability, and requirements for data quality and interpretability. It is shown that isolated use of AI techniques for scaling or scheduling does not yield sustainable benefits in industrial settings, whereas the most robust effects are achieved when intelligent mechanisms are embedded into managed control loops and operate as adaptive but bounded elements of system governance. The study establishes that the effectiveness of AI-driven optimization is determined not by model sophistication, but by the architectural consistency of intelligent components with the control plane, their ability to respect service constraints, and their impact on system stability. The article is intended for researchers and practitioners in distributed systems, cloud and edge computing, and infrastructure architecture concerned with deploying AI-based optimization under production constraints.

high-load distributed systems, AI-driven optimization, cloud-native architectures, Kubernetes, control loop architecture, scheduling, workflow orchestration, production environments

Alharthi, S., Alshamsi, A., Alseiari, A., & Alwarafy, A. (2024). Auto-scaling techniques in cloud computing: Issues and research directions. Sensors, 24(17), 5551. https://doi.org/10.3390/s24175551

Augustyn, D. R., Wyciślik, Ł., & Sojka, M. (2024). Tuning a Kubernetes horizontal pod autoscaler for meeting performance and load demands in cloud deployments. Applied Sciences, 14(2), 646. https://doi.org/10.3390/app14020646

Dakić, V., Đambić, G., Slovinac, J., & Redžepagić, J. (2025). Optimizing Kubernetes scheduling for web applications using machine learning. Electronics, 14(5), 863. https://doi.org/10.3390/electronics14050863

Femminella, M., & Reali, G. (2024). Application of proximal policy optimization for resource orchestration in serverless edge computing. Computers, 13(9), 224. https://doi.org/10.3390/computers13090224

Femminella, M., & Reali, G. (2024). Comparison of reinforcement learning algorithms for edge computing applications deployed by serverless technologies. Algorithms, 17(8), 320. https://doi.org/10.3390/a17080320

Hurtado Sánchez, J. A., Casilimas, K., & Caicedo Rendon, O. M. (2022). Deep reinforcement learning for resource management on network slicing: A survey. Sensors, 22(8), 3031. https://doi.org/10.3390/s22083031

Li, W., Li, X., Chen, L., & Wang, M. (2025). Microservice workflow scheduling with a resource configuration model under deadline and reliability constraints. Sensors, 25(4), 1253. https://doi.org/10.3390/s25041253

Nascimento, B., Santos, R., Henriques, J., Bernardo, M. V., & Caldeira, F. (2024). Availability, scalability, and security in the migration from container-based to cloud-native applications. Computers, 13(8), 192. https://doi.org/10.3390/computers13080192

Tran, M.-N., & Kim, Y. (2025). Hybrid resource quota scaling for Kubernetes-based edge computing systems. Electronics, 14(16), 3308. https://doi.org/10.3390/electronics14163308

Zheng, T., Wan, J., Zhang, J., & others. (2022). Deep reinforcement learning-based workload scheduling for edge computing. Journal of Cloud Computing, 11(1), 3. https://doi.org/10.1186/s13677-021-00276-0