AWS Certified Solutions Architect – Associate (SAA-C03) — Question 418
A company is developing a new machine learning (ML) model solution on AWS. The models are developed as independent microservices that fetch approximately 1 GB of model data from Amazon S3 at startup and load the data into memory. Users access the models through an asynchronous API. Users can send a request or a batch of requests and specify where the results should be sent.
The company provides models to hundreds of users. The usage patterns for the models are irregular. Some models could be unused for days or weeks. Other models could receive batches of thousands of requests at a time.
Which design should a solutions architect recommend to meet these requirements?
Answer options
- A. Direct the requests from the API to a Network Load Balancer (NLB). Deploy the models as AWS Lambda functions that are invoked by the NLB.
- B. Direct the requests from the API to an Application Load Balancer (ALB). Deploy the models as Amazon Elastic Container Service (Amazon ECS) services that read from an Amazon Simple Queue Service (Amazon SQS) queue. Use AWS App Mesh to scale the instances of the ECS cluster based on the SQS queue size.
- C. Direct the requests from the API into an Amazon Simple Queue Service (Amazon SQS) queue. Deploy the models as AWS Lambda functions that are invoked by SQS events. Use AWS Auto Scaling to increase the number of vCPUs for the Lambda functions based on the SQS queue size.
- D. Direct the requests from the API into an Amazon Simple Queue Service (Amazon SQS) queue. Deploy the models as Amazon Elastic Container Service (Amazon ECS) services that read from the queue. Enable AWS Auto Scaling on Amazon ECS for both the cluster and copies of the service based on the queue size.
Correct answer: D
Explanation
Amazon SQS is ideal for decoupling the asynchronous API requests and buffering bursts of incoming messages. Amazon ECS is the correct compute choice because downloading and loading a 1 GB model into memory on AWS Lambda would cause severe cold-start latency, and AWS Auto Scaling cannot be used to scale vCPUs on Lambda functions. By using AWS Auto Scaling with ECS, the system can scale both the task count and the underlying cluster capacity to zero during idle periods, and scale up dynamically based on the SQS queue size during traffic spikes.