DOCSEARCHbyAlgolia
Submit Docs

Azure Architecture Center

repository·main·Indexed 24 days ago

https://github.com/microsoftdocs/architecture-center

Microsoft patterns and practices for designing and implementing cloud solutions on Azure. Includes architectural guidance for AI/ML workloads, such as automated video analysis using Azure AI Video Indexer and Azure Machine Learning, as well as the Baseline Microsoft Foundry landing zone architecture for generative AI.

Tokens
191.1K
Snippets
43
Records
866
Agent score
83%

What's inside Azure Architecture Center

  1. Overview of MLOps v2 architectures

    main

    MLOps v2 provides three distinct Azure architectures designed for end-to-end continuous integration and continuous delivery (CI/CD) and retraining pipelines. These architectures are built using the Azure Machine Learning service and are intended to be deployable, repeatable, and maintainable patterns for AI workloads.

    The three primary architecture types are:

    • Classical machine learning: For tabular structured data.
    • Computer vision (CV): Focused on segmentation and image classification.
    • Natural language processing (NLP): For text-based tasks.

    For implementation details and sample deployment templates, use the Azure MLOps v2 GitHub repository.

  2. Overview of Microsoft security solutions for AWS

    main

    Microsoft security solutions can be used to secure Amazon Web Services (AWS) environments by applying defense-in-depth controls aligned with Zero Trust principles. This approach provides consistent visibility, enforcement, and protection across AWS account access and workloads.

    Key security components include:

    • Microsoft Entra ID: Manages identity, SSO, conditional access, and lifecycle automation.
    • Microsoft Defender for Cloud Apps: Provides visibility, anomaly detection, and session control.
    • Microsoft Defender for Cloud: Offers cloud security posture management (CSPM), attack path analysis (APA), and workload protections.
    • Microsoft Purview: Handles multicloud data discovery, governance, and compliance.
    • Microsoft Sentinel: Provides unified visibility, threat intelligence, and response automation.
    • Microsoft Defender XDR: Delivers real-time containment and cross-platform detection.
    • Microsoft Security Copilot: Assists with incident investigation and automated remediation using natural language.

  3. Overview of Azure load balancing services

    main

    Azure provides several services to distribute workloads across computing resources to optimize usage, maximize throughput, and improve availability. These services vary based on whether they operate at the network layer (Layer 4) or the application layer (Layer 7), and whether they function globally or regionally.

    Key Azure Load Balancing Services

    • API Management: A managed service for publishing, securing, and monitoring HTTP(S) APIs. It can load balance traffic across a back-end pool using round-robin, weighted, or priority-based methods. Note that this is an API gateway first and a general-purpose load balancer second.
    • Application Gateway: A regional Layer-7 proxy load balancer. It provides application delivery controller functionality, including routing, TLS offloading, and Web Application Firewall (WAF) capabilities. It also supports Layer-4 load balancing for TCP and TLS.
    • Application Gateway for Containers: A regional Layer-7 product specifically designed for dynamic traffic management of workloads running in a Kubernetes cluster.
    • Azure Front Door: A global Layer-7 application delivery network. It provides site acceleration, SSL offload, path-based routing, and fast failover for web applications.
    • Load Balancer: A high-performance Layer-4 service handling TCP and UDP traffic. It is designed for ultra-low latency and can be deployed in regional or cross-region topologies.
    • Traffic Manager: A global DNS-based load balancer. It directs traffic at the domain level to services across Azure regions. Because it relies on DNS, failover may be delayed by DNS caching and TTL values.

  4. Overview of Azure Sandbox

    main

    Azure Sandbox is a Terraform-based project that provides a modular and reusable framework for deploying sandbox environments in Azure. It allows developers to quickly provision foundational infrastructure—including virtual networks, virtual machines, and databases—to accelerate experimentation, testing, and development.

    Because a fully provisioned environment can be expensive, the project is designed to be modular: you can deploy only the specific components required for your scenario or manage costs by stopping/deallocating virtual machines when not in use.

  5. Components for measuring application carbon impact

    main

    The following Azure services are used to implement the sustainability measurement architecture:

    • Carbon optimization in Azure: Primary source for measuring and visualizing Azure workload emissions at the resource-group or resource level.
    • Microsoft Cloud for Sustainability API: Provides foundational emissions data and retrieves subscription emissions information.
    • Application Insights: Provides application performance management (APM) data, serving as a proxy for workload utilization and efficiency.
    • Azure Data Lake Storage: Acts as the centralized repository for raw snapshot data used in calculations.
    • Logic Apps: Automates workflows to integrate proxy sources, storage, and calculation systems.
    • Azure Functions: Executes serverless, event-driven code to perform the actual SCI score calculations.
    • Azure Automation: Uses runbooks (PowerShell) to implement complex logic for improving application efficiency.
    • Power BI: Provides the analytics and visualization layer for stakeholder dashboards.

  6. Orchestrate machine learning with Azure Databricks

    main

    This architecture provides a standardized MLOps (Machine Learning Operations) process using Azure Databricks. It is designed to help data scientists and engineers transition machine learning models and pipelines from development to production environments.

    Key characteristics of this solution:

    • MLOps Maturity: Targets level 4 of MLOps maturity.
    • Core Strategy: Uses the promote code that generates the model approach. Instead of promoting a pre-trained model file, this approach focuses on managing and promoting the code used to generate the model, ensuring reproducibility and better version control.
    • Capabilities: Enables full automation, continuous monitoring, and robust collaboration.

  7. Components of the hybrid Azure Files architecture

    main

    The following components are used to build a hybrid file sharing solution:

    • Azure Storage & Azure Files: Azure Storage provides the infrastructure, while Azure Files provides the managed SMB file shares that can be mounted by Windows, Linux, and macOS.
    • Azure Virtual Network: The fundamental building block for the private cloud environment.
    • Azure ExpressRoute: Provides a private, dedicated connection from on-premises to the Microsoft cloud.
    • Azure VPN Gateway: Provides secure connectivity via site-to-site VPN using IPsec and IKE protocols.
    • Azure Private Link & Private Endpoint: Enables private connectivity to Azure PaaS services (like Azure Files) by using a private IP address from your virtual network, avoiding the public internet.
    • Azure Firewall: Can be configured as a DNS proxy to act as an intermediary for DNS requests, simplifying management by removing the need for dedicated DNS virtual machines.

  8. Architecture for an end-to-end stream processing pipeline

    main

    This reference architecture demonstrates how to build a real-time stream processing pipeline using Azure services. The workflow involves:

    1. Data Ingestion: Real-time data streams (e.g., ride and fare information) are sent to Azure Event Hubs.
    2. Stream Processing: An Azure Stream Analytics job reads from the Event Hubs, correlates records from different streams, and performs temporal joins and windowed aggregations.
    3. Data Storage: The processed results are written as JSON documents to Azure Cosmos DB.
    4. Data Analysis: Microsoft Power BI loads data from Cosmos DB for historical analysis, or can receive direct streams from Stream Analytics for real-time visualization.
    5. Monitoring: Azure Monitor tracks performance metrics across all deployed services to ensure solution health.

  9. Identify Azure services for IoT architecture

    main

    Azure provides several core services to support different layers of an IoT architecture:

    • Messaging & Management: Azure IoT Hub provides device-to-cloud messaging and device management.
    • Provisioning: Azure IoT Hub Device Provisioning Service (DPS) handles the provisioning of devices for Azure IoT Hub.
    • Edge-based Data Plane: Azure IoT Operations provides a unified data plane running on Azure Arc-enabled Kubernetes clusters for edge systems.
    • Messaging Protocols: MQTT broker feature in Event Grid provides scalable publish-subscribe messaging using the MQTT protocol.
    • Digital Modeling: Azure Digital Twins enables the digital modeling of physical environments.

  10. Browse the Cloud Design Patterns catalog

    main
    The Cloud Design Patterns catalog provides solutions for common distributed system challenges. Each pattern includes a description of the problem addressed, considerations for application, and Azure-based implementation examples (sometimes including code samples). Patterns are categorized by the Well-Architected Framework pillars they support, such as Reliability, Security, Cost Optimization, Operational Excellence, and Performance Efficiency.

  11. Select an Azure container service

    main

    Azure offers several container services depending on your orchestration needs, management preference, and workload type:

    • Azure Kubernetes Service (AKS): A fully managed Kubernetes service for enterprise-grade security and governance. Best for complex orchestration and full control.
    • Azure Container Apps: A serverless container platform designed for running microservices and containerized applications without managing underlying infrastructure.
    • Azure Container Instances (ACI): A fast, simple way to run containers without orchestration. Ideal for event-driven applications and batch jobs.
    • Azure Red Hat OpenShift (ARO): A fully managed OpenShift service for enterprises requiring Red Hat support and specific platform capabilities.
    • Azure Container Registry (ACR): A private registry service used for building, storing, and managing container images and artifacts.

  12. Monitor hybrid OS workloads with Azure Monitor

    main

    This reference architecture provides a pattern for using Azure Monitor to track the performance and availability of operating system (OS) workloads running in virtual machines (VMs). It supports a hybrid environment including:

    • Microsoft Azure: Azure VMs and Virtual Machine Scale Sets (VMSS).
    • On-premises environments: Main offices and branch offices.
    • Third-party cloud providers: Non-Azure cloud environments.

    The solution centralizes telemetry (logs and metrics) into a Log Analytics workspace, which acts as the administrative boundary and data aggregation point for various Azure Monitor services like Application Insights, Analysis, Visualizations, Alerts, and Autoscale.