Question

An e-commerce company runs an order-processing application on a fleet of EC2 instances in an Auto Scaling group. During flash sales, the frontend web tier generates orders far faster than the backend worker instances can process them, causing dropped requests and failed transactions. The company wants to smooth out the traffic spikes so that no orders are lost even when the backend is temporarily overwhelmed, and it wants the workers to process orders at their own pace. Which architecture change best meets these requirements?

Accepted Answer

Place the order requests into an Amazon SQS queue that the backend worker instances poll, and scale the worker Auto Scaling group based on the queue depth (ApproximateNumberOfMessagesVisible).

Answer

Publish order requests to an Amazon SNS topic that the backend workers subscribe to, so each worker receives every order in real time.

Answer

Add an Application Load Balancer in front of the backend workers and enable connection draining to hold requests until instances are ready.

Answer

Vertically scale the backend worker instances to a larger instance type so they can handle the peak order rate during flash sales.