Which solution meets the requirements and is MOST cost-effective?

By:
On:
In: SAA-C02
Tagged:
With: 0 Comments
A solutions architect is designing the cloud architecture for a company that needs to host hundreds of machine learning models for its users.During startup, the models need to load up to 10 GB of data from Amazon S3 into memory, but they do not need disk access.Most of the models are used sporadically, but the users expect all of them to be highly available and accessible with low latency.

Which solution meets the requirements and is MOST cost-effective?

Deploy models as AWS Lambda functions behind an Amazon API Gateway for each model.

Deploy models as Amazon Elastic Container Service (Amazon ECS) services behind an Application Load Balancer for each model.

Deploy models as AWS Lambda functions behind a single Amazon API Gateway with path-based routing where one path corresponds to each model.

Deploy models as Amazon Elastic Container Service (Amazon ECS) services behind a single Application Load Balancer with path-based routing where one path corresponds to each model.

Explanations:

AWS Lambda is not suitable for models that require large amounts of memory (10 GB) as it has a memory limit of 10 GB, and the models need to be highly available.

Amazon ECS is a suitable option, but deploying a separate ECS service for each model would be inefficient and not cost-effective when most models are used sporadically.

While AWS Lambda can be used for the models, having each model as a separate Lambda function would result in excessive management complexity and inefficiency.

Using Amazon ECS with a single Application Load Balancer and path-based routing allows for efficient and cost-effective management of sporadically used models.

Leave a Reply

Your email address will not be published. Required fields are marked *

1 × 1 =