Design Resilient Architectures

A web front end receives unpredictable bursts of requests that a slower backend cannot process in real time, and no requests may be lost. Which design best decouples the two tiers?

• AHave the front end call the backend synchronously and retry on timeout.
• BScale the backend vertically to a very large instance to absorb every burst.
• CPlace an Amazon SQS queue between the front end and the backend workers.
• DStore requests in a client-side cache until the backend becomes available.

An SQS queue buffers bursts and lets the backend consume work at its own pace with no lost requests. Synchronous retries still overwhelm a slow backend, vertical scaling has limits and a single point of failure, and a client-side cache does not guarantee delivery.

When an order is placed, several independent systems (email, analytics, and inventory) must each receive the event and process it separately. Which service best delivers the event to all of them?

• APublish the event to an Amazon SNS topic that all three systems subscribe to.
• BWrite the event to a single SQS queue that all three systems poll together.
• CCall each system synchronously in sequence from the order service.
• DStore the event in a database table that each system queries on a schedule.

An SNS topic fans a single published message out to all subscribers, so each system receives its own copy. A shared queue delivers each message to only one consumer, synchronous chaining couples the systems, and database polling adds latency and load.