Managing GPU Cost - AKS Cost Management Series - Episode 7 - Azure Kubernetes Service

In this episode, we explore how to manage GPU costs in your Azure Kubernetes service infrastructure.
Kaysie Yu talks to Sachi Desai and explore GPU cost management topics such as picking up the right GPUs, Driver Types, NAP compatibility, Spot vs on-demand and several other GPU related topics.

Related Links
- GPUs on AKS: https://learn.microsoft.com/en-us/azure/aks/use-nvidia-gpu?tabs=add-ubuntu-gpu-node-pool
- GPU Driver Type: https://learn.microsoft.com/en-us/azure/aks/use-windows-gpu#specify-gpu-driver-type-preview
- AMD GPUs - https://aka.ms/aks/amd-gpus
- Node Auto Provision: https://learn.microsoft.com/en-us/azure/aks/node-autoprovision
- GPU VM availability by Azure region: https://azure.microsoft.com/en-us/explore/global-infrastructure/products-by-region/table

Timeline
1:05 GPU cost optimization personas and priorities
2:19 The difference between GPUs and CPUs
4:03 How to pick the right GPU sku
7:00 Grid vs CUDA driver types
10:10 A100 vs H100 vs T4
11:30 GPU skus compatible with Node Autoprovision?
12:24 GPU skus with spot vs on demand
12:50 What security considerations for GPU software components does AKS maintain?
14:45 Where to learn more and get started?

#azure #GPU #NVIDIA #AMD #costmanagement #costsavings #price #kubernetes #aks #microsoftazure

Видео Managing GPU Cost - AKS Cost Management Series - Episode 7 - Azure Kubernetes Service канала Azure Kubernetes Service (AKS)