New: Attribute-based VM selection for Azure Compute Fleet

Today we're excited to announce the preview release of Attribute-based VM selection, a new feature on Azure Compute Fleet enabling you to configure your instance requirements as a set of VM attributes (e.g., memory, vCPU, and storage). These requirements are matched with all suitable Azure VM sizes, simplifying the creation and maintenance of VM configurations. This feature also allows you to seamlessly utilize newer VM generations as they become available and gain access to a wider range of capacity through Azure Spot Virtual Machines. The Compute Fleet service selects and launches VMs that match the specified attributes, eliminating the need to manually choose VM sizes.

Attribute-based VM selection is ideal for scenarios such as stateless web services, large-scale batch processing, big data clusters, or continuous integration pipelines. Workloads like financial risk modeling, log processing, and image rendering can take advantage of the ability to run hundreds of thousands of concurrent cores/instances. When leveraging Spot Virtual Machines, instead of specifying numerous VM sizes and types individually, a simple attribute configuration can now encompass all relevant options, including new ones as they are released. 

Benefits of attribute-based VM selection: 

• Optimal Flexibility for Spot Virtual Machines:  Compute Fleet offer optimal flexibility when deploying Spot Virtual Machines by selecting from a broad range of VM sizes. This aligns with best practices for Spot usage by ensuring flexibility in VM size selection, which improves the likelihood of Azure Spot Virtual Machines meeting and allocating the required compute capacity. 
• Simplified VM Type Selection: With the vast array of VM sizes available, finding the best match for your workload can be a complex task. By specifying VM attributes, Azure VM Scale Sets can automatically select the VM sizes that fulfill your workload’s requirements, simplifying the process and ensuring the right resources are chosen. 
• Automatic Adoption of New VM Sizes: Azure VM Scale Sets can automatically incorporate newer generation VM sizes as they become available. When these new VM sizes meet your specified requirements and align with your chosen allocation strategies, they are seamlessly used by your Scale Set, ensuring your deployment benefits from the latest advancements without manual updates. 

How Attribute-based VM selection Works?

The process involves two main steps: 

1. VM Type Determination: Attribute-based VM selection generates a list of VM types based on specified attributes, chosen Azure Regions, and Availability Zones. 
2. Allocation Strategy Application: The Compute Fleet applies the selected allocation strategy to the generated list. 

For Azure Spot Virtual Machines, Attribute-based VM selection supports both capacity-optimized and lowest-price allocation strategies. For Standard VMs, ABS supports the lowest-price allocation strategy. The Compute Fleet resolves attribute configurations into a list of suitable VM types and initially launches the lowest-priced VM to meet the On-Demand portion of the capacity request, proceeding to the next lowest-priced VM if necessary. 

This feature enables more flexible VM type configurations, eliminating the need for extensive instance-type lists. This way, you can automatically leverage newer VM generations when they are introduced in your selected Region. Additionally, Attribute-Based VM Selection enhances the ability to access more capacity through Spot requests efficiently. 

With Attribute-Based VM Selection, managing VM size configurations becomes easier and more scalable, ensuring that your workloads run smoothly with optimized cost and performance.

Availability & pricing

Attribute-based VM selection is now available in all Azure Public Regions with the exception of those based in China. There is no additional cost for using Attribute-based VM selection.

To learn more, visit the Attribute based VM selection documentation page and sign up to try out the preview features to test these innovations today.

Sign up Form: Compute Fleet preview features