AWS Certified Solutions Architect – Associate (SAA-C03)Design High-Performing ArchitecturesMedium
A data analytics company needs to process petabytes of historical data stored in Amazon S3. The processing jobs are computationally intensive but can tolerate interruptions. The company wants to minimize costs while completing the analysis within a flexible timeframe. Which compute solution should they use?
- ASpot Instances EC2 instances with Amazon EMR.
- BReserved Instances EC2 instances with Amazon EMR.
- CAWS Fargate for Amazon ECS with long-running tasks.
- DOn-Demand EC2 instances with Amazon EMR.
Show answer & explanationAnswer & explanation
Correct answer: A. Spot Instances EC2 instances with Amazon EMR.
Amazon EMR with EC2 Spot Instances is the most cost-effective solution for processing large datasets that can tolerate interruptions. Spot Instances offer significant discounts compared to On-Demand instances, and EMR is well-suited for distributed data processing frameworks like Spark or Hadoop, which can handle retries for interrupted jobs.
Why the other options are wrong
- B. Reserved Instances are for predictable, long-term workloads, not ideal for temporary, interruptible data processing where cost minimization is key.
- C. AWS Fargate is a serverless compute engine for containers, but for petabyte-scale data processing with EMR, EC2 instances (especially Spot) offer more flexibility and cost optimization for this specific use case.
- D. On-Demand instances are more expensive than Spot Instances, making them less cost-effective for interruptible workloads.
EMR with Spot Instances
Running Amazon EMR clusters on EC2 Spot Instances provides a highly cost-effective solution for processing large datasets with distributed frameworks, especially for workloads that are fault-tolerant and can handle interruptions.
- Significant cost savings (up to 90% off On-Demand).
- Ideal for interruptible, fault-tolerant workloads.
- EMR handles distributed processing and job retries.
- Suitable for big data analytics (Spark, Hadoop, Hive, Presto).
- Flexible timeframe for completion.
Memory trick: Spot Instances are like bargain hunters for your EMR jobs, finding the cheapest compute.