/azure-prepare

App Service SKU Selection

SKU Comparison Matrix

Pricing Overview

💡 Tip: Figures are representative for Windows OS in Central US, as of 2026-04. Prices vary by region, OS, and offer. Use the Azure Pricing Calculator for exact figures.

Save by using Reserved Instances and Savings Plans

Cost savings can be made on Premium V3, Premium V4 and Isolated V2 plans by committing to reserved instances for 1 or 3 year terms, details can found at https://learn.microsoft.com/azure/cost-management-billing/reservations/prepay-app-service.

Alternatively cost savings can be made using Azure Savings plans.

Decision Criteria

Production workload?
├─ No → Free (F1) or Basic (B1) for dev/test
└─ Yes
   Need deployment slots, auto-scale, or backups?
   ├─ No → Basic (B1-B3) if budget-constrained (supports VNet integration and Private Endpoints)
   └─ Yes
          Need network isolation (dedicated ASE)?
          ├─ Yes → Isolated (I1v2+)
          └─ No
             Need more than 5 deployment slots or more than 10 instances?
             ├─ Yes → Premium (P1v3+)
             └─ No → Standard (S1-S3) with Private Endpoints and VNet integration

Feature Unlock Summary

Key features unlocked at each tier:

Bicep — App Service Plan with SKU

resource appServicePlan 'Microsoft.Web/serverfarms@2025-03-01' = {
  name: planName
  location: location
  sku: {
    name: 'P1v3'
    tier: 'PremiumV3'
    capacity: 2 // number of instances
  }
  kind: 'linux'
  properties: {
    reserved: true // required for Linux
  }
}

Terraform — App Service Plan with SKU

resource "azurerm_service_plan" "plan" {
  name                = var.plan_name
  location            = azurerm_resource_group.rg.location
  resource_group_name = azurerm_resource_group.rg.name
  os_type             = "Linux"
  sku_name            = "P1v3"
}

Scaling Within a Tier

Scale up (change SKU) vs scale out (add instances):

⚠️ Warning: Scaling from one tier family to another (e.g., Standard to Premium) may cause a brief restart. Schedule changes during low-traffic windows.

Recommendations by Workload