Installation

This guide will show you how to install AeroSim on your local infrastructure.

Important

Make sure to have an NVIDIA GPU compatible with CUDA 12.x, which is required to run the simulations.

Installing on Windows

Enviroment setup

The pre-requirements for installation are:

1. WSL2 (Windows Subsystem for Linux) for Windows (instructions here)

   1. You can Install by running: wsl.exe --install in Prompt

   2. Restart the computer for actions to take place

   3. Create a Ubuntu machine: wsl.exe --install Ubuntu-22.04 in Prompt

   4. Enter the machine: wsl.exe

      • First time it will ask for user and password

      • The password will be asked when you run a command with sudo. Don’t lose it.

2. NVIDIA CUDA Enabled on WSL (instructions here)

   • Download CUDA for WSL2 in NVIDIA

     • Operating system: Linux

     • Architecture: x86_64

     • Distribution: WSL-Ubuntu

     • Version: 2.0

     • Installer type: deb (network) (what you prefer)

   • Copy the commands and paste to the Ubuntu terminal

     • Obs.: if you have keyring problems, delete the old one:

     sudo apt-key del 7fa2af80
     

   • Check installation: nvidia-smi

     • The commands should output GPU informations, drivers and CUDA version

Important

CUDA application have some limitations on WSL2. AeroSim doesn’t guarantee that there won’t be performance impact on the usage of the solver in WSL2.

Linux systems are recommended for avoiding performance issues. Another option is to use a virtual machine (VM) with a Linux distribution, and then proceed as a Linux Installation.

After installing these steps, you can proceed as a Linux installation.

Installing on Linux (Ubuntu)

Enviroment setup

The pre-requirements for installation are:

• NVIDIA GPU with CUDA 12.x support

• NVIDIA GPU driver (with CUDA 12.2 support)

  • GPU driver version must be: >=535

  • Note: in case of WSL2, the driver installed is on Windows, don’t install the GPU driver in the WSL2 Ubuntu.

• Docker installed

  • After installing, make sure docker is running

  sudo systemctl enable docker
  sudo systemctl start docker
  

  • Run this commands, to run docker without root

  sudo groupadd docker
  sudo usermod -aG docker $USER
  newgrp docker
  

• NVIDIA Container Toolkit installed

  • Tool to use CUDA and NVIDIA GPU in Docker

  • After installation:

  sudo nvidia-ctk runtime configure --runtime=docker
  sudo systemctl restart docker
  

Note

CUDA will be installed by docker. It is required only to install the NVIDIA drivers and assert compatibility with CUDA 12.2

After installing, test if everything is alright with:

docker run --rm --gpus all nvidia/cuda:12.2.2-devel-ubuntu22.04 bash -c "nvidia-smi; nvcc --version"

It should show GPU and CUDA informations.

Example of output after successful installation

Note

For other Linux operational systems, as Fedora, CentOS, Debian, etc. similar steps can be taken. The requirements and steps are the same, with only the commands changing from distro to distro.

Installing AeroSim

After installing the pre-requisites, download and install AeroSim with

export AEROSIM_KEY=<key_provided>
sudo systemctl start docker
curl -sSL https://download.aerosim.io/aerosim-solver/install.sh | bash

Acquire the encryption password from an AeroSim representative

After the installation, to run AeroSim solver, just run

aerosim_solver

Now wait a bit and access http://localhost:5050 and use the web interface!

Local usage

The AeroSim local package creates by default a user for login:

• Username: admin

• Password: admin

In order to run simulations, the GPUs must be enabled for usage and queue activated, which can be done in http://localhost:5050/analytics.