Cost Optimization for Event Grid

What is Azure Event Grid and Why Cost Optimization Matters

Azure Event Grid is a fully managed event routing service that enables you to react to events from Azure services and custom sources. It uses a publish-subscribe model where event publishers send events and subscribers (e.g., Azure Functions, webhooks, Event Hubs) process them. While Event Grid is cost-effective for many scenarios, costs can escalate quickly if events are over-produced, filtered improperly, or subscribed to inefficiently. Cost optimization is not just about reducing spend—it ensures your architecture remains scalable, performant, and aligned with business value. By understanding the pricing model and implementing optimization techniques, you can avoid paying for unnecessary event delivery and processing.

Key Cost Drivers in Event Grid

• Number of published events: Every event published to a topic incurs a per-operation cost, regardless of whether it matches any subscription filter.
• Number of event subscriptions: Each subscription triggers delivery attempts and possible retries, adding to operation counts.
• Event delivery attempts and retries: Failed deliveries lead to retry operations, each costing the same as a successful delivery.
• Dead-letter storage: Storing undelivered events in a storage account incurs additional blob storage costs.
• Throughput and scale: High throughput may require premium tiers or advanced features that have separate pricing.

How to Optimize Event Grid Costs

1. Use Event Filtering to Reduce Volume

Event filtering allows you to subscribe only to events that match specific criteria, preventing unnecessary delivery. Filters can be based on event type, subject, or custom data attributes. Use advanced filtering with JSON arrays or string comparisons to narrow down events.

Example: Creating a subscription with subject filtering (Azure CLI)

az eventgrid event-subscription create \
  --name my-filtered-sub \
  --source-resource-id /subscriptions/.../resourceGroups/myRG/providers/Microsoft.EventGrid/topics/myTopic \
  --endpoint-type webhook \
  --endpoint https://myfunction.azurewebsites.net/api/handler \
  --included-event-types Microsoft.Storage.BlobCreated \
  --subject-begins-with /blobServices/default/containers/images/

This subscription only receives BlobCreated events for blobs under the images container. All other blob events are ignored, saving delivery costs.

2. Batch Events for Higher Efficiency

When sending events to a webhook endpoint, Event Grid can deliver multiple events in a single HTTP request if batching is enabled. Batching reduces the number of delivery operations, lowering costs. You can set the maxEventsPerBatch (up to 5000) and preferredBatchSizeInKilobytes (up to 1024 KB).

Example: Creating a subscription with batching (ARM template snippet)

{
  "type": "Microsoft.EventGrid/eventSubscriptions",
  "apiVersion": "2020-06-01",
  "name": "batch-sub",
  "properties": {
    "destination": {
      "endpointType": "WebHook",
      "properties": {
        "endpointUrl": "https://myhandler.example.com/events",
        "maxEventsPerBatch": 500,
        "preferredBatchSizeInKilobytes": 64
      }
    },
    "filter": {
      "subjectBeginsWith": "/blobServices/default/containers/"
    }
  }
}

With batching, a single delivery operation can handle up to 500 events, dramatically reducing the number of billable deliveries.

3. Configure Appropriate Retry Policies and Dead-Lettering

Event Grid automatically retries event delivery based on a default retry policy (exponential backoff up to 24 hours). You can customize the retry policy to reduce retry operations for transient failures. Also, enable dead-lettering to move undelivered events to blob storage after retries are exhausted, avoiding infinite retry costs.

Example: Setting retry policy and dead-lettering (Azure CLI)

az eventgrid event-subscription create \
  --name optimized-retry-sub \
  --source-resource-id /subscriptions/.../resourceGroups/myRG/providers/Microsoft.EventGrid/topics/myTopic \
  --endpoint https://myfunction.azurewebsites.net/api/handler \
  --max-delivery-attempts 3 \
  --event-time-to-live 300 \
  --dead-letter-endpoint /subscriptions/.../resourceGroups/myRG/providers/Microsoft.Storage/storageAccounts/mystorage/blobServices/default/containers/deadletter

This reduces retries to a maximum of 3 attempts within 5 minutes (event-time-to-live in seconds). After that, events are sent to dead-letter storage, stopping further retry costs.

4. Choose the Right Pricing Tier

Azure Event Grid has two tiers: Basic and Premium. Basic is suitable for most scenarios with a per-operation charge. Premium offers higher throughput, dedicated capacity, and advanced features like private endpoints and managed identities. Evaluate your throughput requirements and choose the tier that avoids over-provisioning. For low-volume applications, Basic is cost-effective. For high-volume, predictable workloads, Premium with reserved capacity can reduce per-operation costs.

5. Monitor and Analyze Usage

Use Azure Monitor and Event Grid metrics to track operation counts, delivery failures, and dead-lettered events. Set up alerts to detect anomalies. This data helps you identify subscriptions that generate excessive operations and adjust filtering or batching.

Example: Querying Event Grid metrics via Azure CLI

az monitor metrics list \
  --resource /subscriptions/.../resourceGroups/myRG/providers/Microsoft.EventGrid/topics/myTopic \
  --metric "PublishSuccessCount" \
  --start-time 2025-04-01T00:00:00Z \
  --end-time 2025-04-01T23:59:59Z \
  --interval PT1H \
  --output table

Regular monitoring allows you to correlate cost spikes with specific event types or subscriptions, enabling targeted optimizations.

Best Practices for Ongoing Cost Management

• Design event schemas with cost in mind: Include only necessary properties to reduce payload size (affects bandwidth and storage).
• Consolidate multiple subscriptions into one with advanced filtering: Instead of creating separate subscriptions for each event type, use a single subscription with an array of subject filters.
• Use system topics for built-in Azure services: They are free to create and only charge for operations, similar to custom topics.
• Implement idempotent handlers: If a handler receives duplicate events due to retries, ensure it processes them without side effects. This allows you to safely reduce retry attempts.
• Set up cost budgets and alerts: Use Azure Cost Management to set budgets for Event Grid and receive notifications when spending exceeds thresholds.
• Review subscription usage regularly: Delete unused subscriptions or topics to avoid paying for stale event delivery.
• Use partner topics only when necessary: Partner topics (from SaaS providers) may have separate pricing; evaluate if you need real-time events vs. polling.

Conclusion

Cost optimization for Azure Event Grid is a continuous process that begins at architecture design and extends through monitoring and refinement. By applying event filtering, batching, thoughtful retry policies, and appropriate tier selection, you can significantly reduce the number of billable operations while maintaining reliable event delivery. Coupled with regular monitoring and best practices such as consolidating subscriptions and minimizing payloads, you can ensure that your event-driven architecture remains both performant and cost-efficient. Start small, measure your baseline, and iterate on these techniques to align Event Grid costs with your application’s value.