Windows Server 2025 and Azure Stack HCI 24H2 to improve AI workload mgmt

Microsoft announced upcoming features in Windows Server 2025 and Azure Stack HCI 24H2 to improve support for AI workloads utilizing NVIDIA GPUs.

These updates can provide businesses with more significant control and flexibility when managing AI tasks on their servers. A key point in the announcement is advanced support for NVIDIA GPUs, which allows for tasks like training, learning, and inferencing complex AI models.

The upcoming releases also introduces a Discrete Device Assignment (DDA) functionality within failover clusters. This facilitates assigning entire GPUs to individual virtual machines (VMs), which overall can lead to high availability for critical AI workloads.

A new feature called GPU Partitioning (GPU-P) allows further optimization. GPU-P allows a single physical GPU to be divided into multiple isolated partitions. Each VM in a failover cluster can be assigned its own partition, which maximizes resource utilization, leading to a more cost-effective solution.

Live migration capabilities are included to ensure uninterrupted operation during maintenance or upgrades. VMs with assigned GPU partitions can migrate across cluster nodes without impacting ongoing workloads.

The Windows Administration Center (WAC) serves as a centralized platform for managing virtualized GPUs. Administrators can use WAC to view, configure, and assign GPU partitions to VMs, simplifying management tasks.

Microsoft’s ongoing collaboration with NVIDIA, now the second-most valuable company in the world, ensures these features meet industry security standards. This allows businesses to run AI workloads efficiently.

More here.