Exam AWS Certified Machine Learning - Specialty topic 1 question 21 discussion

AD is correct

CloudTrail is use to track scientist how ofthe they deploy a model CloudWatch for monitoring GPU and CPU so answer is A & D

To monitor SageMaker model deployments, track resource utilization, and log errors, the Machine Learning Specialist should use: AWS CloudTrail – Tracks API activity, such as: Model deployments (e.g., CreateModel, CreateEndpoint) Notebook access and actions SageMaker job executions Amazon CloudWatch – Monitors and logs operational metrics, such as: CPU & GPU utilization of SageMaker endpoints Invocation errors and latencies Custom metrics from deployed models Logs from training jobs and inference endpoints (via CloudWatch Logs)

I think AWS Config is still not the service designed to track how often Data Scientists are deploying models, nor does it track operational performance metrics like GPU and CPU utilization or the invocation errors of SageMaker endpoints. and AWS CloudTrail continues to be the service that will track and record user activity and API usage, which includes deploying models in Amazon SageMaker. So the answers are still A and D - CloudTrail and CloudWatch.

AD is correct

A. YES - to track deployements B. NO - AWS Health is to track AWS Cloud itself (eg. is a zone down ? ) C. NO - AWS Trusted Advisor to give recommendations on infra D. YES - for errors E. AWS Config

I also believe that A and D are correct. Can someone please explain to me the main differences between CloudWatch and CloudTrail? I find the documentation a bit confusing about it

Option E AWS Config to record all resource types, then the new resources will be automatically recorded in your account. Option A CloudTrail is use to track scientist how of the they deploy a model Option D CloudWatch for monitoring GPU and CPU

Log Amazon Sagemaker API Calls with AWS CloudTrail - https://docs.aws.amazon.com/sagemaker/latest/dg/logging-using-cloudtrail.html

I wouldn't be so sure about CloudTrail, AWS Configs also tracks Sagemaker and the resource "AWS::Sagemaker::Model"

A&D CloudWatch and ClouTrail

AD Are Correct.

absolutely

cloudtrail and cloudwatch, no thinking