AWS Certified Machine Learning – Specialty — Question 323

A growing company has a business-critical key performance indicator (KPI) for the uptime of a machine learning (ML) recommendation system. The company is using Amazon SageMaker hosting services to develop a recommendation model in a single Availability Zone within an AWS Region.

A machine learning (ML) specialist must develop a solution to achieve high availability. The solution must have a recovery time objective (RTO) of 5 minutes.

Which solution will meet these requirements with the LEAST effort?

Answer options

• A. Deploy multiple instances for each endpoint in a VPC that spans at least two Regions.
• B. Use the SageMaker auto scaling feature for the hosted recommendation models.
• C. Deploy multiple instances for each production endpoint in a VPC that spans least two subnets that are in a second Availability Zone.
• D. Frequently generate backups of the production recommendation model. Deploy the backups in a second Region.

Correct answer: C

Explanation

Deploying Amazon SageMaker endpoints across multiple instances in a VPC spanning at least two subnets in different Availability Zones provides automatic high availability and failover, easily satisfying the 5-minute RTO with minimal configuration. Multi-region architectures proposed in other options introduce significant deployment complexity and latency, which violates the requirement for the least effort. Auto scaling alone does not protect against an Availability Zone outage unless a multi-AZ infrastructure is already established.