Azure Event Hubs
What it is
Azure Event Hubs is a managed, high-throughput event-streaming platform. It accepts a continuous flow of records—such as device readings, application logs, or website clicks—and lets several consumers read that flow independently.
Think of a river with numbered points along its bank. Many boats add water upstream; each observer can begin at a chosen point and follow the same river at its own pace. An Event Hub is an append-only stream, not a single work item removed by the first worker.
Why it exists
A telemetry or log workload can produce far more small records than a normal request-and-response application can handle one by one. Event Hubs ingests the stream, retains it for a configured time, and lets analytics, monitoring, and archival consumers process it separately.
Key ideas
- A producer sends events to an event hub. A consumer reads events from it.
- The stream is divided into partitions, which enable parallel processing. Events are ordered within a partition, not across all partitions.
- A consumer group is an independent view of the stream. For example, Northwind’s operations dashboard and its long-term analytics job can each read the same telemetry without taking it from the other.
- A consumer tracks its read position, called an offset, and commonly saves a checkpoint after processing. If it restarts, it can resume from that position.
- Retention is time-based and configured by tier. Event Hubs is not a permanent archive; use a suitable storage or analytics destination when long-term retention is needed.
Northwind example
Northwind’s delivery vans send location and temperature readings every few seconds. The vans publish those records to Event Hubs. One consumer group powers a live dashboard for the operations team; another detects temperature anomalies; a third captures the stream for later analysis. A delayed dashboard can catch up from its saved checkpoint while the anomaly service continues independently.
Event Hubs, Event Grid, and Service Bus
| Need | Start by considering |
|---|---|
| Ingest and let several consumers read a high-volume stream of telemetry or logs | Event Hubs |
| React when a discrete state change occurs | [[event-grid |
| Hold and coordinate a business task until a worker processes it | [[service-bus |
The word event appears in all three. The design question is whether consumers need a stream to read, a notification to react to, or a task to complete.
Common confusion
- Event Hubs is not a queue where one successful worker removes a task for everyone else. Different consumer groups can read the same records.
- Partition ordering is useful only when the producer chooses keys and the consumer design respects the partition boundary.
- Reading a record does not prove the business action based on it succeeded. Consumers need checkpoints, retries, and safe repeat processing.
Related
event-grid · service-bus · azure-functions · azure-monitor · northwind-traders-case-study
Sources and re-audit
- Microsoft Learn — What is Azure Event Hubs?, checked 2026-09-16.