If I enable timeSlicing scheduling on the GPU in EKS, will it be affected when applying for resources using Hami scheduling policy

[ollylu]

例如 假设我基于 NVIDIA GPU Operator 把GPU 划分为 10个等分

cat << EOF > nvidia-device-plugin.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: nvidia-device-plugin
  namespace: kube-system
data:
  any: |-
    version: v1
    flags:
      migStrategy: none
    sharing:
      timeSlicing:
        resources:
        - name: nvidia.com/gpu
          replicas: 10

然后 基于 Hami 调度策略 申请资源时，如：

  resources:
    limits:
      nvidia.com/gpu: 2 # 请求 2 个 vGPU
      nvidia.com/gpumem-percentage: 50 

申请到的是 2个vGPU 是物理GPU 还是 按 timeSlicing 划分的 2/10个 GPU ？

[ollylu]

是否会叠加两套调度（驱动时间片 + HAMi 自己的限制），可能产生额外的上下文切换开销？

[wawa0210]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Will two sets of scheduling (driven time slice + HAMi's own limitation) be superimposed, which may incur additional context switching overhead?