Exam DP-100 topic 4 question 10 discussion

I also thought it was D, but it is B for real time and endpoint

Question keyword 'it is important to be able to monitor - the data submitted to the web service and - the predictions the data generates' New GUI is item 'Application Insights diagnostics' with toggle button 'Enabled'. You can practise with sample model Sample PyTorch model https://www.kaggle.com/code/xiedaicheng/download-pytorch-model/output , upload, then create new deployment, then see this option at step 2.

B. To implement a monitoring solution for the deployed model that supports a business-critical application and requires minimal administrative effort, the best option is to enable Azure Application Insights for the service endpoint and view logged data in the Azure portal. Option B, to enable Azure Application Insights for the deployed model, provides a scalable and cost-effective way to monitor the data submitted to the web service and the predictions generated by the data. Azure Application Insights can be easily integrated into Azure Machine Learning and provides powerful analytics tools for tracking and analyzing usage, performance, and errors in real-time. With Azure Application Insights, you can quickly identify and troubleshoot issues, and you can set up alerts to notify you when specific events occur. ACD do not provide real-time monitoring capabilities for the deployed model.

I don't think this can be B because App Insights will now show prediction information, only utilization/traffic etc. The answer has to be D. Review https://learn.microsoft.com/en-us/azure/machine-learning/v1/how-to-use-mlflow?tabs=azuremlsdk#track-runs-running-on-azure-machine-learning

B is the correct answer. Only App Insights can be used here. The question does not talk about MlfLow model. So D is not in question.

On 20Feb2022

D is correct. B provides mostly telemetry data about the service itself.

I think both B and D could work but B is easier to setup