AWS Certified Machine Learning – SpecialtyMachine Learning Implementation and OperationsEasy
An e-commerce company uses an ML model for real-time product recommendations. The model is deployed on a SageMaker endpoint. To ensure high availability and fault tolerance, they need to ensure that the inference endpoint can withstand an Availability Zone (AZ) outage. Which SageMaker endpoint configuration best addresses this requirement?
- ADeploy the model to a single instance with auto-scaling enabled.
- BDeploy the model to a single instance in a private subnet.
- CDeploy the model to multiple instances within a single Availability Zone.
- DDeploy the model to multiple instances across multiple Availability Zones.
Show answer & explanationAnswer & explanation
Correct answer: D. Deploy the model to multiple instances across multiple Availability Zones.
Deploying the model to multiple instances across multiple Availability Zones ensures that if one AZ experiences an outage, the endpoint can still serve traffic from instances in other healthy AZs, providing high availability and fault tolerance.
Why the other options are wrong
- A. A single instance, even with auto-scaling, is vulnerable to a single AZ outage.
- B. A private subnet relates to network access, not high availability across AZs.
- C. Multiple instances within a *single* AZ will all be affected if that AZ goes down.
SageMaker Multi-AZ Endpoint
A SageMaker real-time endpoint configured with instances deployed across multiple AWS Availability Zones to provide high availability and fault tolerance.
- Protects against single AZ outages
- Enhances reliability for critical applications
- Requires specifying instance count and type for each AZ or enabling multi-AZ deployment
- Traffic is automatically routed to healthy instances
Memory trick: Multiple Zones Keep Endpoints Alive.