Dynamic vNPU Scheduling
Function Highlights
Dynamic vNPU scheduling requires Ascend Device Plugin to report the number of available AI Cores on the node where Ascend Device Plugin resides. After a virtualization task is reported, Volcano schedules the task to a node that meets the task requirements. After receiving the request, Ascend Device Plugin of the node automatically splits vNPUs and mounts the task to complete the entire dynamic virtualization process. In this process, you do not need to divide vNPUs in advance, and vNPUs can be automatically reclaimed after the task is complete. This process supports scenarios where your requirements on computing power change continuously.
Before You Start
Before using dynamic vNPU scheduling, you need to learn about the types, allocation rules, and allocation templates supported by Ascend AI processors. For details, see Virtual Instances.
Required Component
- Volcano
- Ascend Device Plugin
- Ascend Docker Runtime
- ClusterD
- NodeD
Instructions
- Refer to Installation and Deployment for component installation.
- Refer to Dynamic vNPU Scheduling (Inference) for feature usage.