Prerequisites
As we reviewed in the overview section, all the end user's compute instances and services will be deployed onto a shared host Kubernetes cluster. Let us review the prerequisites for the host cluster.
Host Kubernetes Cluster¶
Rafay MKS is Rafay's Kubernetes Distribution based on Upstream Kubernetes and it supports a wide variety of configurations and operating systems. We will provision a single node Rafay MKS based Kubernetes cluster for our guide.
Resources¶
Since we will be deploying a few compute instances simulating multiple users, please ensure that the node is resourced adequately i.e.
- CPU: > 4 cores (x64 architecture)
- Memory: > 16GB, and
- Storage: >100GB
Operating System¶
For this exercise, we recommend using a node with Ubuntu 22.04 LTS because this is also supported by Nvidia's GPU Operator software. The OS can either be installed on a bare metal system or a VM.
GPUs¶
Use any Nvidia GPU that is supported by the Nvidia GPU Operator for Kubernetes. You can still proceed with this guide if you do not have access to a GPU by using the GPU Simulator.
Infrastructure Options¶
Select the option that matches resources available to you.
-
Server
Provision on a bare metal or VM based Linux Server with real or simulated GPUs
-
Windows Laptop
Provision on your Windows Laptop with simulated GPUs
-
Mac Laptop
Provision on your Mac Laptop with simulated GPUs