AI-Enabled Datacenter Networking for Cloud-Native Applications: Performance Optimization, Security, and Scalability

Abstract

The pervasive adoption of cloud-native architectures built upon microservices, containers, and Kubernetes orchestration has introduced unprecedented demands on datacenter networking, requiring dynamic service discovery, fine-grained security policies, sub-millisecond latency guarantees, and elastic scalability across heterogeneous multi-cluster environments. Traditional network stacks designed for monolithic application paradigms are fundamentally misaligned with the ephemeral, horizontally scaled, and API-driven nature of cloud-native workloads. This paper presents a comprehensive investigation of AI-enabled datacenter networking for cloud-native applications across three interconnected pillars: performance optimization through intelligent traffic management and data plane acceleration, security through AI-driven zero-trust microsegmentation and anomaly detection, and scalability through predictive auto-scaling and multi-cluster federation. We introduce the Cloud-Native AI-integrated Networking and Security (CNAIS) framework, a unified architecture that embeds machine learning models directly into the networking data path using eBPF (extended Berkeley Packet Filter), implements transformer-based network intrusion detection at line rate, and employs deep reinforcement learning for predictive resource scaling. Evaluation across simulated cloud-native environments comprising 15,000 microservice instances across 5 Kubernetes clusters demonstrates that CNAIS reduces P99 service mesh latency by 59.1%, achieves 98.7% threat detection accuracy with only 0.3% false positive rate, improves resource utilization by 42.3% through predictive scaling, and maintains consistent performance during 10x traffic burst scenarios. Through systematic evaluation against 62 references spanning academic research, cloud provider documentation, and CNCF project reports, this article provides cloud architects and security practitioners with a unified reference for building AI-optimized, secure, and scalable cloud-native datacenter networks.