Async Patterns

Queues, fan-out, retries, webhooks: when and why to decouple work from the request path.

After this path you will be able to

Recognize when synchronous coupling is the root cause of a reliability problem, and correctly apply message queues, fan-out, and idempotent retries to decouple it.

Interview approach for this path

1.Identify the slow or unreliable operation first. Queues solve coupling problems, not speed problems.
2.Name the broker you'd use: SQS for simple delivery, Kafka for replay and multiple independent consumers.
3.State the delivery guarantee: at-least-once means consumers must be idempotent, so say that explicitly.
4.Address fan-out: if multiple services react to the same event, explain how you'd publish once and let each consumer subscribe.
5.Explain your retry strategy and dead-letter queue setup. Without a DLQ, one bad message can block the whole consumer.
6.Confirm the producer and consumer can scale independently, because that is one of the main reasons to use a queue.

Systems in this path

3 total

Concepts reinforced throughout

Message Queues Idempotency Event-Driven Architecture

Up next

Real-Time Systems

Persistent connections, presence, pub/sub fanout, geo-indexed dispatch: designs where seconds matter.

→