Multi-Geo Help documentation
This article describes how you can create a workspace and publish a web service in different Azure regions. The Azure Products by Region page lists regions where Azure Machine Learning is available.
Create a workspace
- Sign in to the Azure Classic Portal.
- Click +NEW > DATA SERVICES > MACHINE LEARNING > QUICK CREATE. Under LOCATION select another region, such as Southeast Asia.
- Select the workspace, and then click Sign-in to ML Studio.
- You now have a workspace in another region that you may use just like any other workspace. To switch among your workspaces, look to the upper right of your screen. Click the dropdown, select the region, and then select the workspace. Everything is local to the workspace region. For example, all of your web services created from a workspace will be in the same region the workspace is located in.
Open an experiment from Gallery
If you open an experiment from Gallery, you can also select which region you want to copy the experiment to.
Web service management
To programmatically manage web services, such as for retraining, use the region-specific address:
Things to note
- You can only copy experiments between workspaces that belong to the same region this way. If you need to copy experiment across workspaces in different regions, you can use the PowerShell commandlet Copy-AmlExperiment to accomplish that. Another workaround is to publish the experiment into Gallery in unlisted mode, then open it in the workspace from the other region.
- The region selector will only show workspaces for one region at a time.
- A free workspace or Guest Access (anonymous) workspace will be created and hosted in South Central U.S.
- Web services deployed from a workspace in Southeast Asia will also be hosted in Southeast Asia.
Ask a question on the Azure Machine Learning forum.