Connecting to datastores
Datastores are the engines where your data resides and provide access to anyone authorized to do so. In most Python examples you see on the internet, there is a connection string that contains the credentials to connect to a database or a blob store. There are a couple of drawbacks associated with this technique:
- The credentials stored within these scripts are considered a security violation, and you can accidentally expose your protected datasets by publishing a script in a public repository such as GitHub.
- You need to manually update all the scripts when the credentials change.
Azure ML allows you to have a single centralized location where you define the connection properties to various stores. Your credentials are securely stored as secrets within the workspace's associated key vault. In your scripts, you reference the datastore using its name and you can access its data without having to specify the credentials. If, at some point...
