Question

A retail company is designing a new order-processing backend on Google Cloud. Order submissions arrive in unpredictable bursts, and each order must be processed exactly once by downstream fulfillment services. During processing spikes, the fulfillment services occasionally fall behind, and the business requires that no order be lost even if a downstream service is temporarily unavailable. The architect must design the messaging layer to meet these reliability and scalability requirements with minimal operational overhead. Which design should the architect choose?

Accepted Answer

Publish orders to a Pub/Sub topic; fulfillment services consume via a subscription with acknowledgment deadlines and dead-letter topics to buffer and retry failed messages.

Answer

Have the order frontend write orders directly to a Cloud SQL table and have fulfillment services poll the table for new rows.

Answer

Send orders over HTTP directly from the frontend to each fulfillment service and retry with exponential backoff in the frontend code.

Answer

Write orders to a Cloud Storage bucket as JSON files and trigger fulfillment services with an hourly scheduled batch job.