How to Create your first GPU Instance
GPU Overview
GPU servers are designed for artificial intelligence, machine learning and complex modeling. They are equipped with high-end GPUs and huge quantity of cores, memory and storage. In short, GPUs are optimized for taking huge batches of data and performing the same operation over and over very quickly.
There are six steps to provision a new GPU server:
- Choose an image
- Choose a region
- Choose a GPU Instance
- Configure storage
- Configure advanced options
- Name and tag the instance
Important: Your SSH public keys are fetched during the boot process.
If you upload them to the management console after your server is booted, they will not be added to your authorized_keys file.
If you do not want the keys to be downloaded during the next boot, execute the following command on your server:
root@c1-X-Y-Z-T:~# echo manual > /etc/init/ssh-keys.override
Requirements
- You have an account and are logged into console.scaleway.com
- You have configured your SSH Key
Creating a GPU Server
In the Compute section of the side menu, click Compute. If you do not have an instance already created, the product presentation is displayed.
Click Create a server. The creation page displays.
Note: GPU instances are currently available on request. Please open a support ticket to request a GPU instance quota for your account.
Choosing an Image
Start by choosing an image for the cloud instance.
You can choose an image from five sources:
OS Images: Choose your favourite basic Linux distribution, ready to be configured by you.
GPU OS: Choose an Ubuntu Bionic (18.04) optimized for Machine Learning. It comes preinstalled with everything you need to fully use your GPU and Anaconda, the most complete python distribution.
InstantApps: We provide an up-to-date list different pre-installed applications.
My images: You can populate your own list of server templates. See also Create your own image
Snapshots: You can recover a server from a previously saved state. See also Backup your data with snapshots
Choosing a Region
You can choose a region in which your instance will be deployed geographically.
Currently we provide the following locations for GPU instances:
- PAR1: Paris, France
Choosing the GPU Instance
Click on the GPU tab to deploy a GPU instance:
Currently RENDER-S instances are available, featuring a Nvidia Tesla P100 16GB, 10 vCores and 400GB SSD storage space.
Configuring Volumes
Each RENDER-S instance comes with a fixed volume size of 400GB.
Configuring Advanced Options
You can configure advanced options of your server, including:
Naming and tagging the instance
Edit the following information about the instance:
- The instances name
- The tag you want to assign to it (optional). Tags let you organize your servers, you can assign any tag to each server.
Starting the Instance
Click the “Create a new Server” button. This action launches the create server action. The instance will be ready soon after.
Logging to your GPU Instance
On a Mac, Windows or Linux computer, you can follow the tutorial on Logging into the Instance.