Free Amazon MLS-C01 Actual Exam Questions - Question 4 Discussion
An ecommerce company has used Amazon SageMaker to deploy a factorization machines (FM)
model to suggest products for customers. The company's data science team has developed two new
models by using the TensorFlow and PyTorch deep learning frameworks. The company needs to use
A/B testing to evaluate the new models against the deployed model.
...required A/B testing setup is as follows:
• Send 70% of traffic to the FM model, 15% of traffic to the TensorFlow model, and 15% of traffic to
the Py Torch model.
• For customers who are from Europe, send all traffic to the TensorFlow model
..sh architecture can the company use to implement the required A/B testing setup?
The TargetVariant header doesn’t handle geo-based routing by itself, so D wouldn’t fully cover the Europe-only TensorFlow traffic. ALB with listener rules seems better to manage that condition, making A more fitting.
It’s A, ALB supports path and host-based routing, so geo-based rules fit better there.
Maybe D makes the most sense here because using production variants inside one SageMaker endpoint simplifies traffic splitting natively. You can easily assign weights for A/B testing between the FM, TensorFlow, and PyTorch models without managing multiple endpoints. Plus, setting the TargetVariant header for Europe traffic to always hit the TensorFlow model aligns perfectly with the requirement.
Options with load balancers like A or C seem more complex and might not handle geo-routing seamlessly without extra workarounds. So, D feels like the cleanest and most integrated approach.
B tbh, using production variants inside a single SageMaker endpoint makes traffic splitting simpler and more native to SageMaker. You can specify weights for each variant directly, which covers the 70/15/15 split well. Then setting the TargetVariant header for Europe users lets you override routing without needing extra infrastructure like ALB. ALBs don’t natively support geo-based routing, so A might overcomplicate things. D is close but doesn’t address traffic override for Europe users as clearly as B does with the header. So B feels like a cleaner, more integrated solution overall.
The key is routing based on location and traffic split. Option A looks right since ALB supports weighted target groups and path-based routing, plus region-based rules. So, A.