Azure Monitor platform metrics

  • 3 minutes

Azure Monitor collects and aggregates metrics from every component of Azure Machine Learning by default. Azure Monitor platform metrics provide a view of availability, performance, and resilience.

Azure Monitor uses the concept of resource types to identify Azure resources. Resource types are also part of the resource ID for every resource running in Azure. For example, one resource type for Azure Machine Learning is Microsoft.MachineLearningServices/workspaces.

Azure Monitor organizes core monitoring data into metrics on resource types, also called namespaces. Metrics and logs are available for various resource types. The metric categories in the Microsoft.MachineLearningServices/workspaces resource are Model, Quota, Resource, Run, and Traffic. Quota information is for Machine Learning compute only. The Run category provides information on training runs for the workspace.

You can use that data to analyze the performance of your Azure Machine Learning environment. For example, if you want to check how many cores a workspace is consuming:

  1. In the Azure portal, open the Azure Machine Learning resource.

  2. On the left menu, expand Monitoring and select Metrics.

  3. On the chart, make sure that Scope is set to the Azure Machine Learning resource. Make sure that Metric Namespace is set to the namespace of the resource. (You might need to select Add metric if no options appear in the graph.)

  4. Under Metric, scroll down to Quota > Total Cores.

    Screenshot of the metrics dashboard in the Azure portal.

  5. Adjust the details of the chart to match your needs, such as time range, time granularity, and aggregation. You can also add more metrics to the same chart for correlation.