Azure Virtual Machine Scale Sets: Storage Configuration

This document provides detailed information on configuring storage for Azure Virtual Machine Scale Sets (VMSS). Proper storage management is crucial for performance, cost-effectiveness, and data durability.

Overview of VMSS Storage Options

VMSS instances can leverage various Azure storage services to meet their specific needs. The primary considerations for VMSS storage include:

• OS Disk: The disk where the operating system is installed.
• Resource Disk: A temporary disk for caching and scratch space, not persisted.
• Data Disks: Additional disks attached to VMSS instances for application data.

OS Disk Types for VMSS

You can choose different types of managed disks for your VMSS OS disks:

• Standard HDD: Cost-effective for development, testing, and non-critical workloads. Offers lower IOPS and throughput.
• Standard SSD: A balance between cost and performance, suitable for web servers, lightly trafficked enterprise applications, and development/test environments.
• Premium SSD: High-performance, low-latency storage ideal for production and performance-sensitive workloads like databases and critical applications.
• Ultra Disk: The highest performance option, offering configurable IOPS and throughput for the most demanding workloads.

The choice of OS disk type impacts boot times, application performance, and cost.

Configuring Data Disks

Data disks can be attached to VMSS instances to store application data. You can specify the size, type, and count of data disks during VMSS creation or update.

Managed Data Disks

Managed disks offer durability and availability. You can attach existing managed disks or create new ones within the VMSS definition.

When defining data disks in your VMSS template (e.g., ARM template or Terraform), you specify properties like:

• lun: Logical Unit Number for the disk.
• createOption: Whether to create a new disk or attach an existing one.
• diskSizeGB: The size of the disk in Gigabytes.
• managedDisk.storageAccountType: The SKU for managed disks (e.g., Standard_LRS, Premium_LRS, StandardSSD_LRS).

Data Disk Caching

Data disk caching can significantly improve read performance. The available caching options are:

• None
• ReadOnly: Optimal for read-heavy workloads.
• ReadWrite: Use with caution, primarily for OS disks.

ReadOnly caching is generally recommended for data disks to avoid data inconsistency issues.

Shared Image Gallery (SIG) and Storage

When using custom images from Azure Compute Gallery (formerly Shared Image Gallery), ensure that the OS disk and any included data disks are properly configured for performance and size.

Example: ARM Template Snippet for Data Disks

"storageProfile": {
    "osDisk": {
        "createOption": "FromImageReference",
        "managedDisk": {
            "storageAccountType": "Premium_LRS"
        }
    },
    "dataDisks": [
        {
            "lun": 0,
            "createOption": "Empty",
            "diskSizeGB": 256,
            "managedDisk": {
                "storageAccountType": "StandardSSD_LRS"
            }
        },
        {
            "lun": 1,
            "createOption": "Empty",
            "diskSizeGB": 512,
            "managedDisk": {
                "storageAccountType": "Premium_LRS"
            }
        }
    ]
}
            

Performance Considerations

IOPS and Throughput

Each disk type offers different IOPS (Input/Output Operations Per Second) and throughput limits. For VMSS with high I/O demands, choose Premium SSD or Ultra Disks. Monitor disk performance metrics in Azure Monitor.

Disk Striping

For extreme performance requirements, consider using disk striping across multiple data disks. This involves configuring your operating system to stripe data across several LUNs, effectively pooling their IOPS and throughput.

Cost Optimization

Selecting the appropriate disk type is essential for cost management. While Premium SSD and Ultra Disks offer superior performance, they come at a higher cost. For workloads that are not performance-intensive, Standard SSD or even Standard HDD can provide significant cost savings. Always assess your workload's I/O patterns and performance needs to make informed decisions.