# Managed Event Services in the Cloud — Event-Driven Architecture & CQRS

Source: https://www.skillbyai.com/en/event-driven-architecture/p-cloud

> Map event-driven patterns to managed cloud services.

## Picking managed building blocks

Cloud providers offer the building blocks so you do not run every broker yourself. **Event streams (logs)**: Apache Kafka as a managed service (Amazon MSK, Confluent Cloud, Aiven and others), Amazon Kinesis, Azure Event Hubs (which also exposes a Kafka-compatible endpoint), and Google Cloud Pub/Sub for global messaging. **Queues and topics for commands**: Amazon SQS and SNS, Azure Service Bus, RabbitMQ (self-run or managed, such as Amazon MQ). **Event routing**: Amazon EventBridge, Azure Event Grid and Google Eventarc route discrete events by rules to functions and services. **Workflow and saga orchestration**: AWS Step Functions, Azure Durable Functions and Logic Apps, Google Workflows, and Temporal (self-hosted or cloud). **CDC**: Debezium, AWS DMS, Google Datastream. The patterns in this course are the same on every platform; managed services change who operates the infrastructure, plus limits, pricing and features such as ordering and retention.

## Pattern-to-service map

A starting point; check each service's ordering, retention and throughput limits.

```text
pattern                         AWS                    Azure                       Google Cloud
------------------------------  ---------------------  --------------------------  -------------------
event stream / replay           MSK, Kinesis           Event Hubs                  Pub/Sub (+ seek), Managed Kafka
command queue / work queue      SQS                    Service Bus queues          Pub/Sub, Cloud Tasks
fan-out pub/sub                 SNS, EventBridge       Service Bus topics, Event Grid  Pub/Sub, Eventarc
saga orchestration              Step Functions         Durable Functions           Workflows
change data capture             DMS, Debezium on MSK   Debezium, Change Feed       Datastream
```

## Read the limits page first

Message size limits, ordering guarantees, maximum retention and delivery semantics differ between services. A design that relies on 30-day replay or strict ordering must be checked against the chosen service before you commit.

**Quiz:** Which managed service type is the natural fit for orchestrating a multi-step saga with compensations?

- [x] A workflow service such as AWS Step Functions or Google Workflows
- [ ] A CDN
- [ ] An object storage bucket
- [ ] A DNS zone

*Answer:* A workflow service such as AWS Step Functions or Google Workflows. Workflow engines persist saga state and handle retries and compensations.