Basic Information
| Field | Details |
|---|---|
| English name | Decoupling & Event-Driven Architecture |
| Full name | Decoupling & Event-Driven Architecture |
| Chinese description | 解耦与事件驱动架构 |
| Japanese description | 疎結合とイベント駆動 |
| Exam frequency | ⭐⭐⭐⭐⭐ |
| Often confused with | 同步调用 |
In one sentence
Decoupling lets producers and consumers change and scale independently while containing local failures behind stable boundaries.
Core purpose
- SQS provides queues, buffering, and load leveling; SNS provides push delivery and fan-out.
- EventBridge routes by event content and rules; Step Functions explicitly orchestrates multi-step workflows.
- Asynchronous processing lets each side run at its own pace, usually with eventual consistency.
Reliable processing
- At-least-once delivery can duplicate messages, so consumers must be idempotent.
- Visibility Timeout should cover processing time; retries and a DLQ isolate failed messages.
- Monitor backlog, age of the oldest message, failure rate, and processing latency.
- Apply backpressure so producers cannot overwhelm consumers indefinitely.
Exam focus
- Backlog or buffer usually points to SQS; fan-out to SNS; event patterns to EventBridge.
- SQS is not broadcast, SNS is not a traditional durable queue, and EventBridge is not primarily backlog storage.
- More retries are not automatically safer; prevent retry storms and poison messages.
Key takeaway
Reliable async = idempotency + retries + DLQ + monitoring; decoupling = independent change + fault isolation.
Related services
Amazon SQS, Amazon SNS, Amazon EventBridge, AWS Lambda, Step Functions, CloudWatch.