You work as a machine learning specialist for a mobile phone carrier network.
[1] [2] Your team of machine learning specialists needs to produce a model that clusters network users by payment plan and real-time geographic region.
Your management team wants to use the model to consider marketing offerings targeted to a customer's billing plan and geographic region in which they spend the most time.
You have created a k-means model, and you need to test variations of the model on real-time customer data.
Which option is considered the best practice for testing model variations as described by your use case?
Click on the arrows to vote for the correct answer
A. B. C. D.Correct Answer: A.
Option A is correct.
The best practice for your use case is to deploy the multiple variants of your model to a single SageMaker HTTPS endpoint.
Where you can direct traffic to each variant as you see fit.
You accomplish this by creating an endpoint configuration that describes all variants of the model.
Option B is incorrect.
The best practice is to deploy the multiple variants of your model to a single SageMaker HTTPS endpoint, not multiple endpoints.
Also, the best practice is to one endpoint configuration that describes all variants of the model, not multiple endpoint configurations.
Option C is incorrect.
The best practice is to one endpoint configuration that describes all model variants, not multiple endpoint configurations.
Option D is incorrect.
The best practice is to deploy the multiple variants of your model to a single SageMaker HTTPS endpoint, not multiple endpoints.
References:
Please see the Amazon SageMaker developer guide titled Deploy a Model in Amazon SageMaker (https://docs.aws.amazon.com/sagemaker/latest/dg/how-it-works-deployment.html#how-it-works-hosting),
The Amazon SageMaker developer guide titled Step 5: Deploy the Model to Amazon EC2 (https://docs.aws.amazon.com/sagemaker/latest/dg/ex1-model-deployment.html#ex1-deploy-model)
In this scenario, the objective is to test variations of a k-means model on real-time customer data. The goal is to cluster network users by payment plan and real-time geographic region to provide targeted marketing offerings to the customers.
To achieve this goal, the best practice for testing model variations is option C: deploy the multiple variants of your model to the same SageMaker HTTPS endpoint, directing a percentage of traffic to each variant of the model, using an endpoint configuration for each variant of the model that describes the model.
Explanation: Option A, which involves deploying multiple variants of the model to the same SageMaker HTTPS endpoint and directing traffic to each variant using an endpoint configuration that describes all variants, is not considered the best practice for testing model variations. This is because directing traffic to multiple variants on the same endpoint can cause issues such as version conflicts, model coexistence issues, and endpoint downtime.
Option B, which involves deploying multiple variants of the model to multiple SageMaker HTTPS endpoints and directing traffic to each variant using an endpoint configuration for each variant, is not the best practice either. This approach could lead to unnecessary costs and overheads as each endpoint needs to be managed separately, and each endpoint requires a separate API call.
Option D, which involves deploying multiple variants of the model to multiple SageMaker HTTPS endpoints and directing traffic to each variant using an endpoint configuration that describes all variants, is not the best practice either. This approach can lead to version control issues, and it can be difficult to manage the multiple variants effectively.
Option C is considered the best practice because it allows the deployment of multiple variants of the model to the same SageMaker HTTPS endpoint. The traffic is then directed to each variant using an endpoint configuration for each variant that describes the model. This approach allows efficient management of the model variations, and it is cost-effective as it requires only one endpoint to be managed. Additionally, this approach allows testing of each model variation under real-time traffic conditions, which can provide valuable insights into the performance and effectiveness of each model variation.
In summary, option C is the best practice for testing model variations in this scenario, as it allows efficient management of model variations, is cost-effective, and allows testing of each model variation under real-time traffic conditions.