Q21: A One-Deal-a-Day Site That Handles Millions of Requests With Millisecond Latency

Problem Info

Domain	Design High-Performing Architectures
Services	Amazon S3, CloudFront, API Gateway, Lambda, DynamoDB
Difficulty	Medium

Question

An ecommerce company is launching a one-deal-a-day website — one product on sale every 24 hours. It must handle millions of requests per hour with millisecond latency during peak traffic.

Which solution meets these requirements with the least operational overhead?

Correct Answer

Host static content in an S3 bucket, front it with a CloudFront distribution (S3 as origin). Build backend APIs with API Gateway + Lambda. Store data in DynamoDB.

Why this is correct

This is the SAA exam’s canonical fully serverless, fully managed stack, and every layer is chosen specifically to avoid capacity planning:

• S3 + CloudFront: the marketing page, images, and other static assets are served from CloudFront’s edge caches — capable of absorbing massive concurrent read traffic with millisecond edge latency, without provisioning a single server.
• API Gateway + Lambda: the backend (e.g., “buy now,” inventory checks) scales automatically with incoming request volume — Lambda spins up as many concurrent executions as needed for a burst, then scales back to zero when idle. No fleet of application servers to size or Auto Scale manually.
• DynamoDB: a fully managed, single-digit-millisecond NoSQL database that scales throughput automatically (with on-demand capacity mode), matching the unpredictable, spiky traffic pattern of a flash-sale product going live.

Every component here is serverless and auto-scaling by default, which is exactly what “least operational overhead” is testing for — no instances, no capacity planning, no patching.

Why the alternatives fall short

• EC2 instances in an Auto Scaling group behind an ALB, with RDS as the database — can technically scale, but Auto Scaling reacts to metrics with some lag, isn’t instantaneous like Lambda’s concurrency model, and requires ongoing instance/AMI/patch management — real operational overhead compared to a fully serverless stack.
• Amazon ElastiCache in front of RDS for read scaling — helps read performance for a relational database, but doesn’t address the core requirement of the backend API layer scaling itself, and still leaves you managing RDS instance sizing and Multi-AZ failover manually.
• Host the entire site on EC2 with pre-warmed instances for the expected traffic spike — “pre-warming” implies manual capacity prediction ahead of each day’s sale, which is itself significant operational effort, and risks under- or over-provisioning if the actual spike differs from the forecast.

Exam Tip

“Millions of requests, unpredictable/spiky traffic” + “millisecond latency” + “least operational overhead” → the fully serverless stack: S3 + CloudFront (static) + API Gateway + Lambda (compute) + DynamoDB (data). This exact five-service combination is one of the most frequently recurring “correct answer shapes” across the entire SAA-C03 exam — recognize it on sight.