What is Amazon SQS?

A queue represents work waiting for a consumer

SQS is useful when a producer should not wait for downstream processing or know which worker performs it. Messages remain in the queue until consumed and deleted or until their retention period expires.

The queue absorbs bursts and exposes backlog as an operational signal. This does not create infinite capacity: producers, consumers, retention, payload size, and downstream throughput still need limits and ownership.

Design around the message lifecycle

Receiving a message does not remove it. The visibility timeout hides it while a consumer works; successful processing must be followed by deletion. If the consumer fails or visibility expires first, the message becomes available again.

Set visibility longer than expected processing with room for variance, or extend it deliberately for long tasks. Make the operation idempotent because redelivery is a normal recovery behavior.

Choose Standard or FIFO from semantics

Standard queues prioritize very high throughput and at-least-once delivery with best-effort ordering. FIFO queues add ordered message groups and deduplication behavior for workloads that require ordered processing within a group.

Operate the backlog, not only the workers

Monitor oldest-message age, visible and in-flight message counts, receive count, consumer errors, throttling, and dead-letter activity. Oldest age often communicates customer delay better than raw queue depth.

Define retry limits and a redrive path for failed messages. A dead-letter queue is an isolation mechanism, not a completed recovery process; someone or something must inspect, correct, and safely replay the work.