Microsoft Docs

Scaling Azure App Service

Azure App Service provides powerful capabilities for automatically scaling your web applications up or out to meet demand. Understanding how to configure and manage scaling is crucial for ensuring performance, availability, and cost-effectiveness.

Understanding Scaling Concepts

There are two primary types of scaling in Azure App Service:

• Scale Up: This involves increasing the resources available to a single instance of your app. This includes CPU, memory, and disk space. You scale up by moving to a higher pricing tier (e.g., from F1 to B1).
• Scale Out: This involves increasing the number of instances running your app. This distributes the load across multiple instances, improving performance and availability. You scale out by increasing the instance count within a given pricing tier.

Automatic Scaling Rules

Azure App Service allows you to configure automatic scaling rules based on various metrics. These rules enable your app to adjust its instance count dynamically without manual intervention.

Common Metrics for Scaling:

• CPU Percentage: Scales out when average CPU usage exceeds a defined threshold.
• Memory Percentage: Scales out when average memory usage exceeds a defined threshold.
• HTTP Queue Length: Scales out when the number of pending HTTP requests increases.
• Disk Queue Length: Scales out when the disk I/O is high.

Configuring Automatic Scaling in the Azure Portal:

1. Navigate to your App Service in the Azure portal.
2. In the left-hand menu, under "Settings," select "Scale out (App Service plan)."
3. Choose between "Manual scale" and "Custom autoscale."
4. For "Custom autoscale," define your rules. You can set minimum and maximum instance counts and specify trigger conditions based on metrics.
5. You can also configure a default capacity for when no rules are met.

Manual Scaling

While automatic scaling is often preferred, manual scaling provides direct control over your instance count. This is useful for predictable traffic patterns or during planned maintenance.

To scale manually:

1. Navigate to your App Service.
2. Go to "Scale out (App Service plan)."
3. Select "Manual scale."
4. Enter the desired number of instances.
5. Click "Save."

Best Practices for Scaling

• Monitor Performance: Regularly review your application's performance metrics to identify bottlenecks and adjust scaling rules accordingly.
• Set Realistic Thresholds: Avoid setting thresholds too low or too high, as this can lead to unnecessary scaling or insufficient capacity.
• Define Max Instances: Always set a maximum number of instances to prevent runaway scaling and control costs.
• Test Your Scaling: Simulate load on your application to test how your scaling rules behave under stress.
• Choose the Right Pricing Tier: Ensure your App Service Plan pricing tier provides sufficient resources for your scaling needs. Scaling out won't help if the individual instances are underpowered.

Example Autoscale Rule

Consider a rule that scales out your app when the average CPU percentage goes above 70% for 10 minutes, and scales in when it drops below 30% for 15 minutes. This helps to maintain performance during peak loads while conserving resources during quieter periods.

// Example configuration snippet (conceptual, not actual API)
{
  "autoscaleSettings": {
    "minCapacity": 1,
    "maxCapacity": 10,
    "defaultCapacity": 2,
    "rules": [
      {
        "name": "ScaleOutOnCpu",
        "description": "Scale out when CPU > 70%",
        "scaleAction": {
          "direction": "Increase",
          "value": "1",
          "count": 1
        },
        "metricTrigger": {
          "metricName": "CpuPercentage",
          "statistic": "Average",
          "timeAggregation": "PT10M", // 10 minutes
          "threshold": 70
        }
      },
      {
        "name": "ScaleInOnCpu",
        "description": "Scale in when CPU < 30%",
        "scaleAction": {
          "direction": "Decrease",
          "value": "1",
          "count": 1
        },
        "metricTrigger": {
          "metricName": "CpuPercentage",
          "statistic": "Average",
          "timeAggregation": "PT15M", // 15 minutes
          "threshold": 30
        }
      }
    ]
  }
}