Common Eventstream Patterns in Fabric

Eventstreams in Fabric enable real-time data ingestion and processing with common patterns for various scenarios.

What Are Eventstreams?

Eventstreams capture and route streaming data in Fabric: - Connect to event sources - Transform in-flight - Route to multiple destinations - Low-latency processing

Source Patterns

Event Hub Integration Connect Azure Event Hubs: - High-throughput ingestion - Partitioned streams - At-least-once delivery

IoT Hub Connection Device telemetry: - Device management - Bi-directional communication - Edge processing

Custom Applications Send events via SDK: - REST API ingestion - Language-specific SDKs - Flexible schemas

Processing Patterns

Fan-Out Route to multiple destinations: - KQL database for analysis - Lakehouse for storage - Alerts for monitoring

Filtering Process subset of events: - Reduce noise - Focus on relevant data - Lower processing costs

Windowing Time-based aggregations: - Tumbling windows (fixed intervals) - Sliding windows (overlapping) - Session windows (activity-based)

Enrichment Add context to events: - Reference data lookups - Calculated fields - Data type conversions

Destination Patterns

Hot Path (KQL Database) Real-time queries: - Immediate analysis - Dashboards - Alerts

Warm Path (Lakehouse) Near-real-time storage: - Historical analysis - Batch processing - ML training data

Cold Path (Archive) Long-term retention: - Compliance - Deep historical analysis - Cost-effective storage

Frequently Asked Questions

What is the maximum throughput for Eventstreams?

Eventstream throughput scales based on your Fabric capacity. Higher capacity SKUs support more events per second. For specific limits, check Microsoft documentation for your capacity size.

Can I replay historical events through Eventstreams?

Eventstreams process real-time data. For replay, store events in Lakehouse or KQL database with full history, then reprocess from storage. Event Hub sources support limited replay from their retention window.