How Do You Reduce Azure Compute Gallery Costs?

Gallery Cost Growth Model
=========================

Month 1:  5 definitions x 1 version  x 3 regions x 2 replicas =   30 managed disks
Month 6:  5 definitions x 6 versions x 3 regions x 2 replicas =  180 managed disks
Month 12: 5 definitions x 12 versions x 3 regions x 2 replicas = 360 managed disks
Month 24: 5 definitions x 24 versions x 3 regions x 2 replicas = 720 managed disks

At 128 GB per disk (Standard_LRS @ ~$5.89/mo per disk):
  Month 1:   $177/mo
  Month 6:   $1,060/mo
  Month 12:  $2,120/mo
  Month 24:  $4,240/mo

Without cleanup, costs grow linearly with each new image version.

Image Version Cleanup Decision Tree
=====================================

[Start] Is this the latest or second-latest version?
  |
  +-- YES --> RETAIN (rollback safety)
  |
  +-- NO  --> Is it older than the retention window (90-180 days)?
        |
        +-- NO  --> RETAIN (within retention window)
        |
        +-- YES --> Does any active VM/VMSS/AVD/pipeline reference this version?
              |
              +-- YES --> RETAIN but right-size target regions
              |
              +-- NO  --> Is there a compliance hold on this version?
                    |
                    +-- YES --> ARCHIVE to Standard_LRS in a single region
                    |
                    +-- NO  --> DELETE

# List all galleries in the subscription
az sig list --query "[].{Name:name, ResourceGroup:resourceGroup, Location:location}" -o table

# List all image definitions in a gallery
az sig image-definition list \
  --gallery-name <gallery-name> \
  --resource-group <rg-name> \
  --query "[].{Name:name, OsType:osType, Publisher:identifier.publisher}" -o table

# List all image versions for a definition with age and replication details
az sig image-version list \
  --gallery-name <gallery-name> \
  --gallery-image-definition <definition-name> \
  --resource-group <rg-name> \
  --query "[].{Version:name, Created:publishingProfile.publishedDate, \
    EOL:publishingProfile.endOfLifeDate, \
    Regions:publishingProfile.targetRegions[].name, \
    ReplicaCount:publishingProfile.targetRegions[].regionalReplicaCount, \
    StorageType:publishingProfile.storageAccountType}" -o table

# Count total replicas across all galleries
az sig list --query "[].name" -o tsv | while read gallery; do
  rg=$(az sig show --gallery-name "$gallery" --query "resourceGroup" -o tsv)
  az sig image-definition list --gallery-name "$gallery" --resource-group "$rg" \
    --query "[].name" -o tsv | while read def; do
    az sig image-version list --gallery-name "$gallery" \
      --gallery-image-definition "$def" --resource-group "$rg" \
      --query "[].publishingProfile.targetRegions[].regionalReplicaCount" -o tsv
  done
done | paste -sd+ | bc

# Azure Automation Runbook: Gallery Image Version Cleanup
# Schedule: Weekly, during maintenance window

param(
    [string]$GalleryName,
    [string]$ResourceGroupName,
    [int]$RetentionDays = 180,
    [int]$KeepLatestVersions = 2,
    [bool]$DryRun = $true
)

Connect-AzAccount -Identity

$definitions = Get-AzGalleryImageDefinition -GalleryName $GalleryName \
    -ResourceGroupName $ResourceGroupName

foreach ($def in $definitions) {
    $versions = Get-AzGalleryImageVersion -GalleryName $GalleryName \
        -ResourceGroupName $ResourceGroupName \
        -GalleryImageDefinitionName $def.Name |
        Sort-Object -Property Name -Descending

    # Always keep the latest N versions
    $candidates = $versions | Select-Object -Skip $KeepLatestVersions

    foreach ($ver in $candidates) {
        $age = (Get-Date) - $ver.PublishingProfile.PublishedDate
        if ($age.TotalDays -gt $RetentionDays) {
            # Check for active VM references
            $refCount = (az graph query -q "Resources \
                | where type =~ 'microsoft.compute/virtualmachines' \
                | where properties.storageProfile.imageReference.id \
                    contains '$($ver.Id)'" --query "count_" -o tsv)

            if ([int]$refCount -eq 0) {
                if ($DryRun) {
                    Write-Output "[DRY RUN] Would delete: $($def.Name) v$($ver.Name) \
                        (age: $([math]::Round($age.TotalDays)) days, refs: 0)"
                } else {
                    Remove-AzGalleryImageVersion -GalleryName $GalleryName \
                        -ResourceGroupName $ResourceGroupName \
                        -GalleryImageDefinitionName $def.Name \
                        -Name $ver.Name -Force
                    Write-Output "Deleted: $($def.Name) v$($ver.Name)"
                }
            } else {
                Write-Output "Skipped: $($def.Name) v$($ver.Name) \
                    (active refs: $refCount)"
            }
        }
    }
}

Over-Replication Example
========================

Configured:  5 regions x 8 replicas = 40 disks per version
Actual:      2 regions actively deploying, max 30 concurrent VMs
Required:    2 regions x 3 replicas  = 6 disks per version
Waste:       34 unnecessary disks (85%)

At 128 GB Standard_LRS (~$5.89/disk/mo):
  Configured cost:  $236/mo per version
  Required cost:    $35/mo per version
  Savings:          $201/mo per version

With 10 active versions: $2,010/mo in unnecessary storage

Replica Sizing Decision Flow
=============================

[Start] For each target region:
  |
  +-- Any VM/VMSS/AVD deployments in this region?
        |
        +-- NO  --> Is this a DR-only region?
        |     |
        |     +-- YES --> Keep 1 replica, Standard_LRS
        |     +-- NO  --> REMOVE region from target list
        |
        +-- YES --> Calculate required replicas:
              |
              +-- VM-based: ceil(concurrent_VMs / 20)
              +-- VMSS-based: concurrent_scale_sets + 1
              |
              +-- Is this a production workload?
                    |
                    +-- YES --> max(calculated, 3) + 1-2 buffer replicas
                    +-- NO  --> max(calculated, 1) (dev/test)

# Update replica count for an image version
az sig image-version update \
  --gallery-name <gallery-name> \
  --gallery-image-definition <definition-name> \
  --gallery-image-version <version> \
  --resource-group <rg-name> \
  --target-regions <region1>=<replica-count>=<storage-type> \
                   <region2>=<replica-count>=<storage-type>

# Example: Reduce from 5 regions/8 replicas to 2 regions/3 replicas
az sig image-version update \
  --gallery-name gallery-prod \
  --gallery-image-definition win11-enterprise \
  --gallery-image-version 1.0.0 \
  --resource-group rg-images \
  --target-regions eastus=3=Standard_LRS westus2=3=Standard_LRS

# Remove a region entirely by omitting it from --target-regions
# (the image version must retain at least one target region)

# Adjust Citrix DaaS machine catalog replica ratio
# Run from Citrix Cloud PowerShell or Remote PowerShell SDK

# View current provisioning scheme settings
Get-ProvScheme -ProvisioningSchemeName "MyCatalog" |
    Select-Object ProvisioningSchemeName,
    MachineCount,
    UseSharedImageGallery,
    SharedImageGalleryReplicaRatio,
    SharedImageGalleryReplicaMaximum

# Update the replica ratio from default 40 to 60
Set-ProvScheme -ProvisioningSchemeName "MyCatalog" \
    -CustomProperties @"
<CustomProperties xmlns="http://schemas.citrix.com/2014/xd/machinecreation">
    <Property xsi:type="StringProperty" Name="SharedImageGalleryReplicaRatio" Value="60" />
    <Property xsi:type="StringProperty" Name="SharedImageGalleryReplicaMaximum" Value="10" />
</CustomProperties>
"@

# Verify the change
Get-ProvScheme -ProvisioningSchemeName "MyCatalog" |
    Select-Object SharedImageGalleryReplicaRatio, SharedImageGalleryReplicaMaximum