Run an Azure Machine Learning Designer training pipeline: You should use mlc_cluster (Machine Learning Compute/Compute Cluster). In Azure Machine Learning, a Compute Cluster is a managed compute infrastructure designed specifically for running training scripts and pipelines (batch workloads). It can scale up to multiple nodes to handle heavy processing and scale down to zero when not in use. While a Compute Instance (nb_server) is meant for a single user's development (notebooks), the Compute Cluster is the standard target for executing Designer pipelines.

Deploying a web service from the Azure Machine Learning designer: You should use aks_cluster (Azure Kubernetes Service). When deploying a model as a real-time web service (inference endpoint) for production, AKS is the appropriate compute target. It provides a scalable, secure environment for hosting the API. Compute Instances are typically used for development and testing, and Compute Clusters are for training/batch inference, not for hosting real-time web services.

Microsoft Learn. (n.d.). What are compute targets in Azure Machine Learning? Retrieved from https://learn.microsoft.com/en-us/azure/machine-learning/concept-compute-target

Relevance: This document details the supported compute targets. It identifies Compute clusters as the target for "Training, Batch inference" and Azure Kubernetes Service (AKS) for "Real-time inference."

Microsoft Learn. (n.d.). Tutorial: Designer - Train a regression model. Retrieved from https://learn.microsoft.com/en-us/azure/machine-learning/tutorial-designer-automobile-price-train-score

Relevance: In the "Create a pipeline" section, the tutorial explicitly instructs users to create or select a compute cluster target to run the pipeline.

Microsoft Learn. (n.d.). Deploy a model to an Azure Kubernetes Service cluster. Retrieved from https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-azure-kubernetes-service

Relevance: Confirms that AKS is used for high-scale production deployments of web services.