Running Neurodesktop on Google Cloud (with GPU)

Neurodesktop (https://neurodesk.org) is a flexible, scalable, and browser-based data analysis environment for reproducible neuroimaging. It provides a fully featured destktop in a container allowing users to consume containerised tools enabling researchers to easily create and reproduce analysis.

Neurodesktop can be run in any environment that can run Docker but the real power comes when it can make use of large CPU, memory or GPUs to reduce the analysis time.

In this blog post, I will walk through how to set up a Google Cloud virtual machine to run Neurodesktop with an attached GPU.

First Create a Google Compute Engine (GCE) VM with a connected GPU

Choose or create a new project

Navigate to Compute Engine -> Instances

Select “CREATE INSTANCE”

Configure the Instance

Edit the name of the instance

Choose the Region where you want to run the instance

Choose the GPU option for Machine Configuration

Configure the GPUs

Choose the GPU Type and Number

Note that to the right you can see an estimate of the monthly cost as you choose the configuration.

Also note that depending on the particular GPU chosen, you may need to change zone and/or region to one where those GPUs are available.

Configure the CPU

Choose the CPU Type and number, making sure you have at least 7.5GB RAM

Note that if you cannot find a predefined combination of CPU and memory that is suitable, you can select custom and specify the combination you need.

Configure the O/S and Disk Size

Choose Ubuntu Linux

Version 20.04 LTS x86_64

With 64GB storage (or more if you need it)

Configure Firewall to allow HTTP Traffic

Select “Create”

Wait for the VM to be ready

Configure the instance to run NeuroDesktop

Hit the SSH button to open an SSH Connection to the VM

Install Docker

    sudo apt-get install -y    apt-transport-https     ca-certificates     curl     gnupg-agent     software-properties-common

    curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add –

    sudo add-apt-repository    “deb [arch=amd64] https://download.docker.com/linux/ubuntu \

       $(lsb_release -cs) \

       stable

    “

    sudo apt-get update -y

    sudo apt-get install -y screen docker-ce

    sudo systemctl enable –now docker

    sudo systemctl status docker

Install NVidia Drivers for the GPU

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin

sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600

wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda-repo-ubuntu2004-11-8-local_11.8.0-520.61.05-1_amd64.deb

sudo dpkg -i cuda-repo-ubuntu2004-11-8-local_11.8.0-520.61.05-1_amd64.deb

sudo cp /var/cuda-repo-ubuntu2004-11-8-local/cuda-*-keyring.gpg /usr/share/keyrings/

sudo apt-get update

sudo apt-get -y install cuda

Check that the Nvidia Drivers are installed and working properly

    nvidia-smi

This should show the GPU as visible

Install nVidia Container Toolkit

    distribution=$(. /etc/os-release;echo $ID$VERSION_ID)       && curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg –dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg       && curl -s -L https://nvidia.github.io/libnvidia-container/$distribution/libnvidia-container.list |             sed ‘s#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g’ |             sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

    sudo apt-get update -y

    sudo apt-get install -y nvidia-docker2

    sudo systemctl restart docker

Reboot

    sudo reboot

Reconnect

Wait a short while and hit the reconnect button

Check that the Nvidia Drivers are installed and working properly

    sudo docker run –rm –gpus all nvidia/cuda:11.0.3-base-ubuntu20.04 nvidia-smi

This should, once again, show that the GPU is visible (this time inside Docker)

Run Neurodesktop Docker Container

sudo docker run   –shm-size=1gb -d –privileged –name neurodesktop-20221012 –restart unless-stopped -v ~/neurodesktop-storage:/neurodesktop-storage   -e HOST_UID=”$(id -u)” -e HOST_GID=”$(id -g)” –gpus all -p80:8080 -p 3390:3389 -h neurodesktop-20221012 vnmd/neurodesktop:20221012

Return to the Google Cloud Console

Wait a minute for the Docker container to start then hit the external address on the VM Instance

Test the environment

A new Tab will open with the login screen

Login with the default username: “user” and password: “password”

Open the Desktop

Open a terminal window

Give permissions to the shared folder

    sudo chmod 777 /neurodesktop-storage

Check we have access to the gpu

    nvidia-smi