Event Grid and Event-Driven Architecture

In an event-driven architecture, components communicate by publishing and subscribing to events rather than calling each other directly. A component (producer) emits an event when something noteworthy happens — a file is uploaded, an order is placed, a device sends a reading. Other components (subscribers) react to events they care about, independently and asynchronously. This decouples producers from consumers, improving scalability, resilience, and maintainability.

Azure Event Grid is a fully managed event routing service that delivers events from sources (publishers) to handlers (subscribers) using a push model. Event Grid is designed for reactive, discrete events — something changed and you need to react immediately. It guarantees at-least-once delivery with automatic retry, supports filtering so subscribers only receive relevant events, and can route events to Azure Functions, Logic Apps, webhooks, Event Hubs, and Service Bus queues.

Publishers send events to an Event Grid topic. Subscribers create event subscriptions on a topic, specifying an endpoint and optional filter rules. A single topic can have multiple subscriptions — each subscription receives an independent copy of matching events. Azure services such as Blob Storage, Resource Manager, Service Bus, and Azure Container Registry are built-in event sources with system topics that require no additional configuration.

Events published to Event Grid follow a standard JSON schema with mandatory fields: id (unique identifier), eventType (e.g., Microsoft.Storage.BlobCreated), subject (path to the event resource), eventTime (ISO 8601 timestamp), dataVersion, and data (event-specific payload). Event Grid also supports CloudEvents schema (CNCF standard) for interoperability with other event platforms.

Event subscriptions support filtering to reduce noise at the subscriber end. You can filter by event type (only receive BlobCreated, not BlobDeleted), subject prefix or suffix (only blobs in a specific container), or advanced filters on any field in the event data using operators like StringContains, NumberGreaterThan, and BoolEquals. Filtering happens server-side, so subscribers only receive events matching their criteria.

Three Azure services handle events and messages — choose based on your scenario. Event Grid is for reactive event routing (low volume, discrete, reactive — e.g., resource change notification). Event Hubs is for high-throughput event streaming (millions of events/second, telemetry, big data pipelines). Service Bus is for enterprise messaging with ordering, deduplication, dead-lettering, and transactions (order processing, financial transactions).

Publish custom events to an Event Grid topic using a simple HTTP POST request with the topic's access key. Any service or application that can make HTTP requests can publish events. This makes it trivial to emit events from on-premises applications, third-party services, or Azure services that don't have native Event Grid support. Batch up to 1 MB of events per POST for efficiency.

If an event delivery attempt fails (subscriber returns non-2xx HTTP), Event Grid retries using an exponential backoff with jitter for up to 24 hours (configurable up to 72 hours) with a maximum of 30 retries. After exhausting retries, Event Grid can dead-letter undelivered events to an Azure Blob Storage container for manual investigation. Configure dead-lettering when reliable event delivery is critical and you need to audit or reprocess failed events.

Azure Event Hubs is a distributed data streaming platform capable of receiving and processing millions of events per second. It uses a partitioned consumer model — events are distributed across partitions, and each consumer group reads events independently at its own pace. This enables multiple consumers to process the same stream without coordination. Common use cases include IoT telemetry ingestion, application log aggregation, clickstream analysis, and real-time dashboard data pipelines.

Azure Service Bus is an enterprise messaging broker providing queues (point-to-point) and topics with subscriptions (publish-subscribe). Unlike Event Grid (fire-and-forget) and Event Hubs (streaming), Service Bus guarantees ordered delivery (FIFO queues), duplicate detection, dead-letter queues for processing failures, message sessions for grouped processing, and transactions — all essential for financial and order processing workflows.

A common Azure event-driven pattern: a user uploads a file to Blob Storage, which fires a BlobCreated event to Event Grid. Event Grid routes the event to an Azure Function that processes the file (resizes an image, extracts text, validates data) and writes the result to Azure SQL Database. If processing fails, the event is dead-lettered to a storage container. This entire pipeline requires zero polling and no persistent compute when idle.

Test your understanding of Microsoft Azure Fundamentals (AZ-900) concepts from this lesson.

In this lesson you learned: Azure Event Grid routes discrete events from publishers to subscribers using push delivery with filtering and dead-lettering, the key difference between Event Grid (reactive events), Event Hubs (streaming), and Service Bus (reliable enterprise messaging), and how to build event-driven pipelines by chaining Blob Storage events through Event Grid to Azure Functions. Next up we explore Azure DevOps services.