Function Highlights
Description
The Ascend virtual instance can virtualize an Ascend AI processor (NPU) of a physical machine or virtual machine into several virtual NPUs (vNPUs) and mount the vNPUs to containers. Such virtualization method allocates and reclaims resources of different specifications in a unified manner, allowing users to allocate and release resources repeatedly.
The Ascend virtual instance allows multiple users to share one server and to allocate vNPU resources as needed, making the NPU computing power more accessible and affordable. To be specific, users can share NPU resources on the same server and use containers to isolate resources. This ensures the stability and security of the operating environment. In addition, resources can be allocated and reclaimed in a unified manner, making it easier to manage multiple tenants.
Basic Principles
Ascend NPU hardware resources include AI Cores (used by AI models for computation), AI CPUs, and memory. The Ascend virtual instance virtualizes an NPU into vNPUs based on user requirements. Each vNPU corresponds to a certain number of AI Cores, AI CPUs, and memory resources. For example, if you only need to use four AI Cores, the system creates a vNPU that allocates four AI Cores from the NPU to your container. For details about the Ascend virtual instance scheme, see Figure 1.
