Launch a Data Science Service on Oracle Cloud

Who Should Read This: Cloud engineers, data scientists, data analysts, or machine learning engineers getting started with Oracle Cloud Data Science.

Steps to Launch Oracle Data Science Service:

One of the benefits of cloud data science is the ability to automate the deployment of data science resources and tools. Oracle Cloud makes it easy to deploy and manage data science environments with the Oracle Data Science Service. Oracle Data Science Service comes with Conda, Jupyter Lab, Python 3x, pre-built Conda packs for many different machine learning tasks, and data science templates for many different requirements. Additionally, Oracle Data Science comes with a model catalog, a feature store, jobs, an SDK, and many other things.

To quickly get started, here is what you need to do…

A. Deploy a Stack:

1. Create a compartment to house your data science resources.

Engineering Tip: Don’t put all your resources and services in one compartment. Structure compartments in such a way that allows you or your organization to manage cloud resources effectively. By dedicating a compartment or sub-compartment to Oracle Data Science Service, I can easily manage those compartments costs, resources, and access. I can also remove or change resources in the compartment without risk to resources in other compartments on my cloud.

2. Open Resource Manger on Oracle Cloud console. This can be found quickly by searching for ‘Stacks’ or ‘Resource Manager’. Resource Manager uses Terraform to automate many types of deployments on Oracle Cloud.

3. Select ‘Create Stack’ on the Stacks page. The web from for creating a stack will open. Work through this form to deploy the data science service. This only needs to be done once. After that you can deploy an infinite number of data science projects and notebooks.

4. Select ‘Templates’. This will open a sub-menu of pre-built Terraform templates provided by Oracle for deployment. Data Science is listed under the Services tab. Select the Data Science radial and choose ‘select template’. The stack configuration for Oracle Data Science will now show on the main page for ‘Create Stack’.

5. Provide a name and description.

Engineering Tip: Use the description field to capture notes about your resources. This will help with management later.

6. Ensure the correct compartment is selected. You can move resources to other compartments after they are deployed, but it is easier to deploy it to the correct department to begin with.

7. Add tags and click Next.

Engineering Tip: Utilize OCI tags to better track resources and costs in your tenancy. I find it useful to create a tag for my username in free text to make it easier for me to track the resources I deploy.

8. IAM group and policy configuration will deploy as part of the Terraform script. This saves so much time! Input names as desired.

Engineering Tip: If this is the first deployment, accept the provided defaults. Use common and meaningful names that can be remembered and found easily later.

9. Enable vault. This is optional, but I highly recommend enabling ‘Vault’ support. Vault is used to store secrets and manage encrypted resources.

10. Click Next and review the information.

11. Click the box to ‘Run apply’ and then click ‘Create’. Provided there are no errors, the job will run and complete on its own.

12. If the job succeeds, then you are ready to create a project and deploy a notebook.

Common cause of stack job failure:

One common reason for a data science Terraform job to fail is if any of the configuration variables are already deployed with the same names.

What happened to the policies that were created?

The Terraform script lays down the required policies for you. To review the policies, go to the policies page. Select DataScienceServicePolicies or whatever name you used. You should see a list of policies created for you.