SAA-C03 Question #597: Real Exam Question with Answer & Explanation

Question

A company is developing machine learning (ML) models on AWS. The company is developing the ML models as independent microservices. The microservices fetch approximately 1 GB of model data from Amazon S3 at startup and load the data into memory. Users access the ML models through an asynchronous API. Users can send a request or a batch of requests. The company provides the ML models to hundreds of users. The usage patterns for the models are irregular. Some models are not used for days or weeks. Other models receive batches of thousands of requests at a time. Which solution will meet these requirements?

Options

• ADirect the requests from the API to a Network Load Balancer (NLB). Deploy the ML models as
• BDirect the requests from the API to an Application Load Balancer (ALB). Deploy the ML models
• CDirect the requests from the API into an Amazon Simple Queue Service (Amazon SQS) queue.
• DDirect the requests from the API into an Amazon Simple Queue Service (Amazon SQS) queue.

Explanation

For irregular, asynchronous, and resource-intensive machine learning microservices that need to load significant data at startup, using an Amazon SQS queue with Amazon ECS on AWS Fargate provides scalable, cost-effective, and operationally efficient execution.

Common mistakes.

• A. Deploying on EC2 Spot Instances with an NLB is problematic because Spot Instances can be interrupted, forcing the model to reload 1GB of data and causing service disruptions. NLB is also a Layer 4 load balancer, not typically used for HTTP API routing that an ML model might expose.
• B. Deploying on EC2 On-Demand Instances, even with an ALB, means paying for instances during long idle periods, which is inefficient for irregular usage patterns. Scaling from zero can also be slow when sudden spikes occur due to the 1GB data loading time for each new instance.
• C. While AWS Lambda functions are good for serverless and irregular usage with an SQS queue, loading 1GB of data at startup for each cold start can be slow and may exceed Lambda's optimal performance characteristics for very large payloads or long-running initializations, despite increased memory limits. Fargate offers more control over container startup and resource allocation for such heavy initialization tasks.

Concept tested. Serverless containerization for irregular, data-intensive microservices

Reference. https://docs.aws.amazon.com/AmazonECS/latest/developerguide/what-is-fargate.html

No community discussion yet for this question.