AI in Cloud-Native Architecture: Rethinking Software Design for Data-Intensive and Machine Learning Systems
DOI:
https://doi.org/10.15662/IJEETR.2026.0805007Keywords:
Artificial intelligence, machine learning systems, data engineering, cloud-native architecture, distributed systems, MLOps, data pipelines, model servingAbstract
The migration of computing workloads from on-premises data centers to elastic cloud platforms has changed how data is collected, processed, and turned into intelligence. Cloud-native architecture, which treats elasticity, distribution, automation, and failure as first-class design inputs, has become the default foundation for modern data platforms and machine learning systems. Yet AI and data workloads differ in important ways from the web and transactional applications for which cloud-native practice was first developed. They move and transform large volumes of data, depend on long-running training jobs and scarce accelerators, serve models whose behavior is determined by data rather than code alone, and can degrade silently as the world changes. This paper examines cloud-native design from the perspective of AI and data systems and makes four contributions. First, it reinterprets the foundational principles of the approach, including statelessness, loose coupling, and designing for failure, for data pipelines, feature management, and model lifecycles. Second, it surveys the recurring patterns used to build distributed data and AI systems, from logs and feature stores to streaming pipelines, online inference, and resilience strategies for model serving. Third, it examines how training, inference, and continuous data pipelines reshape assumptions about state, consistency, and resource elasticity. Fourth, it analyzes the trade-offs that architects of AI and data platforms must reconcile, since no single design optimizes for scalability, data correctness, model quality, cost, and operational simplicity at once. The central argument is that building AI on the cloud is not a matter of deploying models on cloud infrastructure. It requires treating data, features, and models as first-class architectural concerns, governed by the same principles of decoupling, versioning, and resilience that cloud-native design applies to services.
References
1. Mell, P. and Grance, T. The NIST Definition of Cloud Computing. NIST Special Publication 800-145, 2011.
2. Gilbert, S. and Lynch, N. “Brewer's Conjecture and the Feasibility of Consistent, Available, Partition-Tolerant Web Services.” ACM SIGACT News, 2002.
3. Burns, B., Grant, B., Oppenheimer, D., Brewer, E., and Wilkes, J. “Borg, Omega, and Kubernetes.” ACM Queue, 2016.
4. Newman, S. Building Microservices. O'Reilly Media.
5. Nygard, M. Release It! Design and Deploy Production-Ready Software. Pragmatic Bookshelf.
6. Kleppmann, M. Designing Data-Intensive Applications. O'Reilly Media, 2017.
7. Dean, J. and Ghemawat, S. “MapReduce: Simplified Data Processing on Large Clusters.” Proceedings of the USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2004.
8. Zaharia, M., et al. “Resilient Distributed Datasets: A Fault-Tolerant Abstraction for In-Memory Cluster Computing.” Proceedings of the USENIX Symposium on Networked Systems Design and Implementation (NSDI), 2012.
9. Kreps, J., Narkhede, N., and Rao, J. “Kafka: A Distributed Messaging System for Log Processing.” Proceedings of the NetDB Workshop, 2011.
10. Akidau, T., et al. “The Dataflow Model: A Practical Approach to Balancing Correctness, Latency, and Cost in Massive-Scale, Unbounded, Out-of-Order Data Processing.” Proceedings of the VLDB Endowment, 2015.
11. Sculley, D., et al. “Hidden Technical Debt in Machine Learning Systems.” Advances in Neural Information Processing Systems (NeurIPS), 2015.
12. Breck, E., Polyzotis, N., Roy, S., Whang, S. E., and Zinkevich, M. “Data Validation for Machine Learning.” Proceedings of Machine Learning and Systems (MLSys), 2019.
13. Gama, J., Žliobaitė, I., Bifet, A., Pechenizkiy, M., and Bouchachia, A. “A Survey on Concept Drift Adaptation.” ACM Computing Surveys, 2014.
14. Dean, J., et al. “Large Scale Distributed Deep Networks.” Advances in Neural Information Processing Systems (NeurIPS), 2012.
15. Moritz, P., et al. “Ray: A Distributed Framework for Emerging AI Applications.” Proceedings of the USENIX Symposium on Operating Systems Design and Implementation (OSDI), 2018.
16. Crankshaw, D., et al. “Clipper: A Low-Latency Online Prediction Serving System.” Proceedings of the USENIX Symposium on Networked Systems Design and Implementation (NSDI), 2017.
17. Kwon, W., et al. “Efficient Memory Management for Large Language Model Serving with PagedAttention.” Proceedings of the ACM Symposium on Operating Systems Principles (SOSP), 2023.
18. Lewis, P., et al. “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks.” Advances in Neural Information Processing Systems (NeurIPS), 2020.
19. Gebru, T., et al. “Datasheets for Datasets.” Communications of the ACM, 2021.
20. Mitchell, M., et al. “Model Cards for Model Reporting.” Proceedings of the Conference on Fairness, Accountability, and Transparency (FAT*), 2019.
21. Dehghani, Z. Data Mesh: Delivering Data-Driven Value at Scale. O'Reilly Media, 2022.
22. Huyen, C. Designing Machine Learning Systems. O'Reilly Media, 2022.
23. Kraska, T., Beutel, A., Chi, E. H., Dean, J., and Polyzotis, N. “The Case for Learned Index Structures.” Proceedings of the ACM SIGMOD International Conference on Management of Data, 2018.
