SAP-C02 Question 369
Select 3A company is migrating its on-premises data warehouse to Amazon Redshift. The solution architect has been tasked with ensuring that the new architecture can handle unpredictable spikes in query workloads while minimizing costs. Additionally, the company wants to integrate the new system with machine learning models hosted on Amazon SageMaker for predictive analytics. Which combination of steps should the architect take to meet these requirements?
- A
Use Amazon Redshift RA3 nodes with managed storage to scale storage and compute independently.
- B
Enable Amazon Redshift concurrency scaling for managing unpredictable query spikes.
- C
Configure Reserved Instances for Amazon Redshift to minimize costs during unpredictable workloads.
- D
Integrate Amazon Redshift with Amazon SageMaker using Redshift ML for predictive analytics.
- E
Use Amazon DynamoDB as the primary data store for unpredictable query workloads.
Show answer and explanation
Correct answers: A, B, D
Explanation
The correct solution involves using Amazon Redshift RA3 nodes to optimize cost and scalability, enabling concurrency scaling to handle unpredictable spikes in query workloads, and integrating with SageMaker using Redshift ML for predictive analytics. Reserved Instances are not suitable for unpredictable workloads, and DynamoDB is not designed for data warehousing or analytical workloads.
- A. Correct.
Correct: RA3 nodes allow you to scale storage and compute independently, which is cost-effective for varying workloads.
- B. Correct.
Correct: Concurrency scaling enables Amazon Redshift to handle unpredictable spikes in query workloads by automatically adding transient capacity.
- C. Incorrect.
Incorrect: Reserved Instances are not suitable for unpredictable workloads due to their rigid cost structure and lack of scalability.
- D. Correct.
Correct: Redshift ML enables seamless integration with SageMaker for predictive model inference without data movement.
- E. Incorrect.
Incorrect: DynamoDB is a NoSQL database and does not serve as a data warehouse or support the analytical workloads required in this scenario.