Azure Messaging and Events: Service Bus, Event Grid, Event Hubs and Storage Queues Compared
Message or Event?
- A message carries data that a sender wants processed by a consumer (an order to process). Delivery and processing matter
- An event says that something happened (a file was uploaded). The sender does not care who reacts
- A stream is a high-volume sequence of events or telemetry
The Four Services
| Service | Model | Best for |
|---|---|---|
| Storage Queue | Simple queue, up to 64 KB messages, huge volume | Basic decoupling, cheap, no ordering guarantees |
| Service Bus | Enterprise broker: queues and topics/subscriptions | Orders, payments, workflows needing reliability, ordering, transactions |
| Event Grid | Push-based event routing (pub/sub) | React to Azure resource events and custom events, serverless triggers |
| Event Hubs | Big data streaming (Kafka-compatible) | Telemetry, logs, clickstream with millions of events per second |
Service Bus
- Queue: point-to-point, one consumer receives each message
- Topic with subscriptions: publish/subscribe, each subscription gets a copy and can filter messages with rules
- Features: dead-letter queue for failed messages, sessions (FIFO per session id), duplicate detection, scheduled delivery, transactions, message TTL, peek-lock receive mode
- Premium tier gives dedicated capacity, VNet/private endpoints, larger messages and zone redundancy
bash
az servicebus namespace create -g rg-int -n sb-shop-482 -l centralindia --sku Standard
az servicebus queue create -g rg-int --namespace-name sb-shop-482 -n orders --max-delivery-count 5 --enable-dead-lettering-on-message-expiration true
az servicebus topic create -g rg-int --namespace-name sb-shop-482 -n order-events
az servicebus topic subscription create -g rg-int --namespace-name sb-shop-482 --topic-name order-events -n billing
python
from azure.servicebus import ServiceBusClient, ServiceBusMessage
from azure.identity import DefaultAzureCredential
client = ServiceBusClient("sb-shop-482.servicebus.windows.net", DefaultAzureCredential())
with client.get_queue_sender("orders") as sender:
sender.send_messages(ServiceBusMessage('{"orderId": 1001}'))
with client.get_queue_receiver("orders") as receiver:
for msg in receiver.receive_messages(max_message_count=5, max_wait_time=5):
print(str(msg))
receiver.complete_message(msg) # remove it. Otherwise it returns after the lock expires
Peek-lock The receiver locks a message, processes it, then completes it. If it fails or times out the message reappears, and after max delivery count goes to the dead-letter queue.
Event Grid
- Routes events from sources (Blob Storage, resource groups, Event Hubs, custom topics) to handlers (Functions, Logic Apps, webhooks, queues)
- You create an event subscription with filters (event type, subject prefix)
- Delivers with retry and dead-lettering. Cheap and serverless, pay per event
bash
az eventgrid event-subscription create --name on-upload \
--source-resource-id $(az storage account show -n stshopprod482 -g rg-data --query id -o tsv) \
--endpoint-type azurefunction --endpoint /subscriptions/<sub>/resourceGroups/rg-app/providers/Microsoft.Web/sites/func-thumbs/functions/MakeThumb \
--included-event-types Microsoft.Storage.BlobCreated --subject-begins-with /blobServices/default/containers/uploads
Event Hubs
- Ingests millions of events per second. Events are stored in partitions for a retention period (default 1 day, more with higher tiers)
- Consumer groups let several apps read the same stream independently, each with its own position
- Capture automatically writes events to Blob or Data Lake
- Supports the Kafka protocol so Kafka apps can connect without code changes
- Process with Stream Analytics, Spark, Functions or Data Explorer
Decision Table
| Scenario | Choose |
|---|---|
| Process orders reliably, with retries and dead-letter | Service Bus queue |
| Send one event to several systems with filtering | Service Bus topic or Event Grid |
| Run a function when a blob is created | Event Grid |
| Ingest IoT telemetry at massive scale | Event Hubs |
| Simple, cheap queue for background jobs | Storage Queue |
| Strict first-in first-out per customer | Service Bus sessions |
| Existing Kafka producers | Event Hubs (Kafka endpoint) |
Practice
- Create a queue and send then receive messages, make one fail until it reaches the dead-letter queue
- Create an Event Grid subscription for blob creation
- Choose the service for five scenarios from the table
Conclusion
Use Service Bus for reliable business messages, Event Grid for event reactions, Event Hubs for high-volume streaming and Storage Queues for simple jobs. Decoupling with messaging makes systems more resilient and scalable.
Quick Summary: message vs event; four services; Service Bus; Event Grid; Event Hubs; decision table.

