Knowledge Graph — Coursera Notes › Organizations › Microsoft › Azure

Azure Kubernetes Service (AKS)

service · part of Azure

A managed Kubernetes service for deploying and scaling containerized applications.

In the context of Azure Machine Learning, AKS is used for production-level deployment of machine learning models, providing high availability and scalability. It is ideal for large-scale, production-ready environments. Deploying to AKS involves registering a model, creating an environment, writing an entry script, creating an inference config, provisioning the AKS cluster, and deploying the web service. AKS clusters typically require sufficient CPU quota (e.g., 12 vCPUs by default) and can be configured with SSL/TLS for encrypted communication.

Model profiling helps determine the CPU and memory resources needed for a deployed service. It is optional but can be time-consuming (up to 25 minutes per model). The results inform the resource allocation for the AKS cluster.

Enabling SSL/TLS on an AKS cluster encrypts data in transit, indicated by a lock icon in the browser URL. This prevents third-party interception of data sent to and from the cluster.

ACI is suitable for lightweight, low-risk deployments like internal testing, while AKS is for production-level scalability and high availability. A typical workflow: deploy to ACI for validation, then to AKS for production.

Inside Azure Kubernetes Service (AKS) (5)

• AKS Deployment Steps — Register the trained model (e.g., sklearn-regression-model.pkl).
• AKS Web Service Testing — After deployment, test the web service using the run method with sample data, or via raw HTTP request.
• AKS Workspace Setup — To deploy models, you need an Azure Machine Learning workspace.
• Azure Container Registry (ACR) — A managed Docker registry service for storing and managing container images, used to push images for AKS or App Services deployment.
• Entry Script (score.py) — The entry script is a Python file (e.g., score.py) that defines how the model is loaded and how predictions are made.

Connections

• Used for Scaling
• Used for Azure Machine Learning Service
• Uses Azure Container Registry (ACR)
• Alternative to Azure Functions
• Alternative to Azure Machine Learning Service
• Used for Scalability
• Alternative to Azure ML Service
• Alternative to Azure App Services
• Related to Azure Container Registry (ACR)
• Related to Azure Machine Learning Service
• Related to Azure Container Instances (ACI)
• Used for Docker container
• Uses MLflow

Also known as: Azure Kubernetes Service (AKS), Kubernetes Service (AKS)

This is the text view of an interactive 3D knowledge graph — open this page with JavaScript enabled to explore it visually.