1. GPU-Accelerated Computing for Deep Learning on Exoscale


    To set up a GPU-accelerated computing environment for deep learning on Exoscale using Pulumi, we need to create a virtual machine (VM) with a GPU attached. Below is an example of a Pulumi program that provisions such an environment. It will include the necessary resources, such as:

    • A security group to define firewall rules allowing SSH access.
    • A SSH key pair for secure access to the VM.
    • A virtual machine instance with a GPU for accelerated computing.

    Make sure to replace the placeholder values (like the SSH public key, the instance size, template, etc.) with your actual values. For the sake of this program, we'll assume you've already registered an SSH key with Exoscale and have the details required for setting up a VM. If not, you'll need to do this before the program can successfully create resources.

    Here is the detailed program written in Python, which uses Pulumi with the Exoscale provider to create the deep learning environment:

    import pulumi import pulumi_exoscale as exoscale # Create a security group to allow SSH access to the VM security_group = exoscale.SecurityGroup('deep-learning-sg', description="Allow SSH access", # Define rules to allow SSH (port 22) ingress from anywhere ingress=[ exoscale.SecurityGroupIngressArgs( description="SSH", protocol="tcp", start_port=22, end_port=22, cidr_list=[""], ), ], ) # Assuming you have an SSH key already registered with Exoscale, specify its name here ssh_key = "your-ssh-key-name" # Create a virtual machine for deep learning vm_instance = exoscale.ComputeInstance("deep-learning-vm", name="deep-learning-instance", zone="YOUR-ZONE-HERE", # Replace 'YOUR-ZONE-HERE' with an appropriate zone where GPU instances are available type="GPU", # Choose a VM type that includes a GPU, this placeholder 'GPU' should be replaced with an actual GPU offering template_id="YOUR-TEMPLATE-ID-HERE", # Replace with the appropriate template ID for a deep learning image ssh_key=ssh_key, security_group_ids=[security_group.id], # Attach the security group created earlier disk_size=100, # Define the disk size, in GB ) pulumi.export("vm_public_ip", vm_instance.public_ip) # Output the public IP of the VM to connect to it later

    In this program, we:

    1. Define a security group that allows inbound SSH connections. The SSH rule permits access from any IP address (, but you should limit it to known IPs for better security.
    2. Assume you have a registered SSH key in Exoscale. You’ll reference it by its name.
    3. Create a GPU-enabled VM instance in a specified zone, with a given Exoscale template appropriate for deep learning (with pre-installed CUDA drivers, for example). Be sure to select a type that includes a GPU, and provide the correct template ID that Exoscale offers for such instances.
    4. Finally, the program exports the public IP of the VM so that you can SSH into it and start using the deep learning environment.

    Remember to replace placeholders like YOUR-ZONE-HERE, GPU, YOUR-TEMPLATE-ID-HERE, and your-ssh-key-name with real values based on your Exoscale account and the resources you've set up.

    To run this Pulumi program, you'll need to have the Pulumi CLI installed and configured, along with Python. If you haven't done so already, you'll need to set up the Exoscale provider configuration with your API credentials. With everything installed and configured, you simply run pulumi up, and it will provision the resources defined in the above program.