Requirements

The MOSTLY AI Synthetic Data Platform is designed to run on Linux systems (Red Hat 7.7 and above or Ubuntu 18.04 or 20.04) with at least 90 GB of space available.

You’ll also need to have the following minimum recommended partition sizes:

  • 30 GB for /var

  • 30 GB for /home

  • 30 GB for /tm

The installation requires a dedicated IP Address available where the users will be able to access the application (more details about its use in the installation instructions below).

The hardware requirements depend on the size of the datasets to be processed; but the next setup can be considered as a good starting point:

Data size

up to 100k subjects and 100 columns

CPU

32 Cores

Memory

128 GB

Disk Storage (SSD)

500 GB

By default, MOSTLY AI will use ports 8080, 8090, 8081 and 5672 for the web interface, Keycloak, Coordinator, and RabbitMQ respectively. It is important that these four ports are open (or that other four ports are open and dedicated to the tool) so MOSTLY AI can operate normally. The ports can be reassigned during the installation.

Installation guide
You can find further technical requirements in the Technical specifications section.

Dependencies

As a containerized application, it is necessary to have Docker Engine (20.10 and above) and Docker Compose (v2.5.0 and above) already installed on the instance where MOSTLY AI will be installed, as they are necessary for running the system. You can follow their official documentation to install or upgrade them according to your operating system:

In the case of air-gapped instances with no internet access, the required dependencies for running Docker and Docker Compose are as follows, and due to interdependencies, they have to be installed in that order:

  • containerd.io-1.6.4

  • docker-ce-cli-20.10.9 + docker-scan-plugin-0.12.0

  • docker-ce-20.10.9 + docker-ce-rootless-extras-20.10.9

  • docker-compose-plugin-2.5.0

Example: RHEL 7.9

Considering the example of a RHEL 7.9 instance, once the packages (rpm) have been downloaded and transferred to the air-gapped host, they can be installed as follows:

Installation guide
sudo yum --disablerepo=docker-ce-stable install -y containerd.io-1.6.4-3.1.el7.x86_64.rpm

sudo yum --disablerepo=docker-ce-stable install -y docker-ce-cli-20.10.9-3.el7.x86_64.rpm docker-scan-plugin-0.12.0-3.el7.x86_64.rpm

sudo yum --disablerepo=docker-ce-stable install -y docker-ce-20.10.9-3.el7.x86_64.rpm docker-ce-rootless-extras-20.10.9-3.el7.x86_64.rpm

sudo yum --disablerepo=docker-ce-stable install -y docker-compose-plugin-2.5.0-3.el7.x86_64.rpm
  • Start docker:
    sudo systemctl start docker

  • Verify Docker Compose:
    docker compose version

Installation guide

System administrator requirements

MOSTLY AI requires an administrative user for installation which has one of the following attributes:

  • member of the docker group, OR

  • has sudo rights for docker, OR

  • is root OR

  • has sudo rights

Installation

  • Download the latest version of the App and Agent and transfer them to the instance where they will be installed:
    mostly-ai-r2.4.1.sh
    mostly-agent-r2.4.1.sh

  • Make them executable:
    chmod +x mostly-*.sh

  • Create a shared folder for the App and the Agent:
    mkdir mostly-data

  • Install the App:
    sudo ./mostly-ai-r2.4.1.sh

The installer will ask for the IP address where MOSTLY AI will be accessed from (public IP address); for the ports, you can just hit enter and the installer will leave the default ports. For Data path, use the previously created folder, e.g.:
/home/ec2-user/mostly-data

Installation guide
  • Launch the installation of the Agent using the same IP address of the instance (private IP address):
    sudo ./mostly-agent-r2.4.1.sh

For the ports, you can just hit enter and the installer will leave the default ports. For Data path, use the previously created folder, or the folder destined for this task, e.g.:
/home/ec2-user/mostly-data

Installation guide

Testing

Log into the app, activate the license with the Customer Success Engineer from MOSTLY AI (offline activation), and run a smoke test with a sample dataset.

Installation guide