workspaces
Summary
Details
Plan for a multi-regional deployment of Azure Machine Learning and associated resources
Impact: High Category: Disaster Recovery
APRL GUID: a86ed26a-59d9-47bd-b440-6bc71b843978
Description:
Ensure you have a recovery strategy defined. Check regional availability and paired regions. Machine Learning doesn't have auto failover. Therefore, you must design a strategy that encompasses the workspace and all its dependencies, such as Key Vault, Azure Storage, and Container Registry.
Potential Benefits:
Provides multi-region disaster recovery strategy
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query
Deploy Azure Machine learning workspace in secondary region
Impact: High Category: Disaster Recovery
APRL GUID: 675d249a-9486-45e3-8e89-863f5802782d
Description:
If your primary workspace is unavailable, switch to the secondary workspace to continue work. Azure Machine Learning doesn't auto-submit jobs to the secondary workspace during an outage. Update your code configuration to point to the new workspace resource.
Potential Benefits:
Provides recovery from regional outages
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query
Ensure to create Machine Learning Compute resources in secondary region
Impact: High Category: Disaster Recovery
APRL GUID: 13794a63-8d95-47ce-acbd-5925ede5b208
Description:
Create compute resources for training a Machine Learning model in selected regions. Ensure both regions have sufficient compute quota for dynamic scaling. Customers must configure HA across zones for attached compute resources like AKS, Azure Databricks, Container Instances.
Potential Benefits:
High availability and disaster recovery
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query
Selecting regions for BCDR, ensure that both regions offer adequate compute quotas
Impact: High Category: Disaster Recovery
APRL GUID: 6e4f0fd1-1853-4b94-9736-6d6d239d2694
Description:
When selecting regions for BCDR, ensure that both regions offer adequate compute quotas to meet your requirements in the same SKU. This ensures that you can failover to the secondary region without any issues.
Potential Benefits:
Provide enough compute resources to the secondary region
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query
Choose SKUs with longer terms and avoid those nearing retirement
Impact: Medium Category: Service Upgrade and Retirement
APRL GUID: 6e2af91f-477d-46a5-b8ce-6cd1b8176550
Description:
When choosing SKUs, opt for those that support longer terms and steer clear of any SKUs that are nearing retirement. This ensures that you can continue to use the SKU for a longer period of time.
Potential Benefits:
supportability, longer term support
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query
Avoid NC and NC_Promo series Azure VMs for machine learning quotas; migrate to newer versions
Impact: High Category: Service Upgrade and Retirement
APRL GUID: cf2569bb-1cf2-46ce-8885-d742dc6f4a4c
Description:
Avoid selecting NC and NC_Promo series Azure virtual machines for machine learning VM quotas and make sure to migrate to newer versions.
Potential Benefits:
Avoid service disruption, longer term support
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query
Make Azure Machine Learning quota requests through the Azure Machine Learning Studio
Impact: High Category: Other Best Practices
APRL GUID: 48ea6480-6263-40ba-8937-326d790e63f6
Description:
Requests for additional Azure Machine Learning quota should be made through the Azure Machine Learning Studio instead of the subscription level in the Azure portal.
Potential Benefits:
Scalability,capacity planning
Learn More:
ARG Query:
Click the Azure Resource Graph tab to view the query