An order processing application has a web front end that accepts orders and a backend fleet that processes them. During flash sales, order volume spikes far faster than the backend can scale, causing failed orders. What architecture change best resolves this?
A) Have the web front end call the backend directly via synchronous HTTP requests with retry logic.
B) Place an SQS queue between the front end and backend; the front end pushes messages to the queue, and the backend consumes them at its own pace.
C) Increase the backend Auto Scaling Group’s maximum capacity to a very high number.
D) Use an SNS topic to broadcast each order to all backend instances simultaneously.
Correct Answer: B
Explanation: Introducing an SQS queue decouples the producer (front end) from the consumer (backend). Orders are durably stored in the queue even if the backend is temporarily overwhelmed, and the backend processes messages at a sustainable rate — smoothing out traffic spikes without dropping orders.
Why the others are wrong: (A) synchronous calls with retries still fail or time out if the backend can’t keep up. (C) raising capacity limits helps but doesn’t solve the fundamental coupling problem, and Auto Scaling still takes time to react. (D) SNS fan-out delivers each message to every subscriber — the wrong pattern for distributing discrete work items across a worker pool.