Microsoft adds cross‑cluster networking, multi‑NIC pod capability, and deeper observability to AKS, aiming to boost performance and manageability for AI and high‑performance computing tasks.
Microsoft Azure Kubernetes Service (AKS) now supports cross‑cluster networking and multi‑NIC pods, delivering the bandwidth and observability needed for AI and high‑performance computing (HPC) workloads.
Cross‑cluster networking
AKS can now connect pods across separate clusters using Azure Virtual Network peering, allowing workloads to communicate with low latency and consistent security policies. This feature simplifies the architecture of distributed AI training jobs that span multiple clusters.
Multi‑NIC pod support
Pods can be assigned more than one network interface, enabling direct access to high‑throughput NICs for data‑intensive tasks. Multi‑NIC pods help reduce bottlenecks in GPU‑accelerated training and large‑scale simulations.
The new capabilities are complemented by deeper observability tools, including enhanced metrics and tracing for network traffic, helping operators pinpoint performance issues quickly.
Benefits for AI and HPC
AI researchers can now distribute model training across clusters without sacrificing network performance, while HPC engineers gain finer control over data movement between compute nodes.
- Reduced latency for inter‑cluster communication
- Higher throughput via multiple NICs per pod
- Integrated monitoring for network health
These updates align AKS with other leading Kubernetes services that offer advanced networking features, positioning Azure as a strong platform for next‑generation compute workloads.
For more details on the networking enhancements, see the Microsoft Azure blog on AKS networking updates.
Explore related coverage on the Infrastructure section and read about Microsoft’s broader cloud strategy on Microsoft on Momentum.