Quick Answer

The answer is three design choices: using Amazon EMR with Spot Instances for task nodes, employing Amazon S3 Select to retrieve only necessary data subsets, and configuring AWS Glue with a scheduled crawler while selecting a lower number of Data Processing Units (DPUs) for smaller jobs. These choices optimize costs for big data analytics pipelines by leveraging the fault-tolerant nature of EMR task nodes to safely use low-cost Spot Instances, reducing data transfer and processing volume through server-side filtering with S3 Select, and matching Glue’s compute capacity precisely to job requirements to avoid over-provisioning. On the SAA-C03 exam, this question tests your ability to apply cost optimization design choices across different AWS services, often appearing as a multi-select scenario where distractors include using On-Demand instances or full data scans. A common trap is assuming all cost savings require sacrificing performance, but here stateless task nodes and selective data retrieval maintain throughput. Memory tip: “Spot, Select, and Scale Down” — Spot for fault-tolerant nodes, Select for filtered data, and Scale Down DPUs for small jobs.