MLOps Reference Architecture deployed on AWS

This repository contains Terraform configuration which serves as an example of how to deploy MLOps architecture on an EKS cluster.

Other repositories

Prerequisites

To use this configuration, you must have Terraform installed at version 0.14 or newer.

You must also have an AWS account.

ATTENTION: Applying this configuration will result in AWS billing you for the provisioned resources.

Installation

Edit my_vars.tfvars file. Set values for the variables mentioned there.
cd into the root of this repository.
Set up access to your AWS account, best by setting environment variables.
Run terraform apply -var-file my_vars.tfvars. Review the plan that Terraform produces.
If you are okay with the plan, answer yes. Terraform will provision the cluster and install MLOps tools in it.

Accessing the tools

The tools are configured in the most basic way. For details on how to change the configuration, please refer to the MLOps reference architecture documentation.

Jupyterhub

The default password is a-shared-secret-password.

Provisioned resources

Please refer to this repository's Readme to learn what resources will be provisioned.

An S3 bucket will be provisioned as a storage backend.

MLOps tools will be installed in the cluster.

Structure

Two Terraform modules are used:

mlops-architecture-eks-cluster to provision a Kubernetes cluster using EKS service of AWS,
mlops-architecture to install MLOps tools in the cluster.

Please refer to the documentation of the modules for the details.

datarevenue-berlin / OpenMLOps-AWS