Exam DP-100 topic 4 question 36 discussion

For Endpoint type the Managed Online Endpoint is the right answer. I was not convinced but then I read this: https://learn.microsoft.com/en-us/azure/machine-learning/concept-endpoints-online?view=azureml-api-2#managed-online-endpoints-vs-kubernetes-online-endpoints

Endpoint type: Kubernetes online Deployment component: Azure Kubernetes Service (AKS) cluster

answers are correct: A Docker image can be used for real-time inferencing of ML models, depending on the deployment solution and the requirements. For example, you can use a Docker image to create a web service that exposes your model via a REST API and responds to requests with predictions in near real-time1. Alternatively, you can use a Docker image to deploy your model to a cloud platform such as Amazon SageMaker2 or Azure Machine Learning3 that provides managed online endpoints for real-time inferencing. However, using a Docker image for real-time inferencing may also introduce some challenges, such as ensuring the compatibility and security of the image, synchronizing the image with the online feature store, and scaling the image to handle the traffic and latency demands4. Therefore, you should carefully evaluate the trade-offs and best practices of using a Docker image for real-time inferencing of ML models.

Deployment component asnwer might be docker image! Because: https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-online-endpoints?view=azureml-api-2&tabs=azure-cli#define-the-deployment Define the deployment: A deployment is a set of resources required for hosting the model that does the actual inferencing. To deploy a model, you must have: (among others) - An environment in which your model runs. The environment can be a Docker image with Conda dependencies or a Dockerfile.

both managed online endpoints and AKS-based online endpoints can be used to deploy machine learning models for real-time inferencing in Azure Machine Learning. Managed online endpoints are fully managed by Azure Machine Learning and provide a simple and cost-effective way to deploy models for real-time inferencing. They also provide out-of-the-box monitoring and logging powered by Azure Monitor and Log Analytics, which includes key metrics and log tables for endpoints and deployments. On the other hand, AKS-based online endpoints provide more flexibility and control, but require more user responsibility to set up and manage. So the only reason I would go for managed online is the mentioning of Azure Log Analytics directly. Second one should be Azure Kubernetes Service (AKS) for deployment. SaM

To enable endpoint and deployment level cost estimates, you can use managed online endpoints. Managed online endpoints work with powerful CPU and GPU machines in Azure in a scalable, fully managed way. They take care of serving, scaling, securing, and monitoring your models, freeing you from the underlying infrastructure management. https://learn.microsoft.com/en-us/azure/machine-learning/concept-endpoints-online?view=azureml-api-2#managed-online-endpoints-vs-kubernetes-online-endpoints

https://learn.microsoft.com/en-us/azure/machine-learning/concept-endpoints?view=azureml-api-2 managed online can provide deployment level cost estimates

I would go for EndpointType: managed online, Deployment component: AKS cluster

Was on exam 2023-03-27

Is it correct?