Skip to content
Justin's Digital Garden

Azure Open Source Day 2023

Agenda

Section titled “Agenda”
  • See app-building demos using Azure and the latest in open-source technologies, cloud-native architectures, and microservices.
  • Tips and best practices for open source from industry leaders at companies like HashiCorp, GitHub, and Redis.
  • Learn to build cloud-native apps for relational and nonrelational data with Azure Cosmos DB, now supporting PostgreSQL.
  • Discover new capabilities in IaaS, PaaS, containers, and serverless computing, including Azure Kubernetes Service (AKS).

Introduction

Section titled “Introduction”

Open source advantages:

  • Distributed global development and many developers can participate
  • Choice in people, software, collaboration
  • Open standards
  • Vendor neutral and competing companies work together
  • AI assisted software development, though disadvantage is security problems like spam pull requests, intrusion use

Demonstration - Open Source + AI to Build an Intelligent App

Section titled “Demonstration - Open Source + AI to Build an Intelligent App”

Pet spotter app, connect owners with their lost pets called “Petspotr”

  • Use ML classification model to identify pet and allows found pets to match to their owner’s pictures

Walk Through

Section titled “Walk Through”

  • Visual Studio Code - extension for k8s

    • Bridge to k8s - manage k8s

  • Bicep: Infrastructure as Code

    • Bicep is open source, has extension fro AKS
    • Azure key vault, k8s, cluster sizes and configuration
      • Containers
      • Azure Cosmos Redis DB
    • AKS cluster built using Bicep configuration file
    • Bicep assists with specification of resources with Intellisense, telling you required fields and values
    • Push bicep to git, GitHub action deploys infrastructure
    • K8s
      • import kubernetes
      • Configures all k8s resources
      • Use dapr for backend configuration

  • Kubernetes (k8s)

    • Distributed Application Runtime (Dapr)
      • Dapr manages complexity of microservices
      • Dapr client can
        • Help programs communicate with state store
        • Handle redis provisioning
        • Publish / subscribe to topics
        • Run programs like front/back end locally in containers
      • Dapr dashboard shows it takes care of use of cloud resources when deployed

  • Web application: report lost pets and found pets

  • Azure load testing - for CI/CD

  • Playwright - cross-browser testing for web apps and has extension for VS Code

  • Machine learning

    • Using huggingface.co ML model, use existing models, image classification
      • Why “huggingface” - interest in AI recognizing emotions like a hug
      • Azure ML can access Hugging Face foundational models with popular models instead of large one time setup (e.g. GPT 2)
    • Azure ML
      • Configure pipeline for input/outputs, pytorch
      • Python, set inputs of pet metadata
      • Build pipeline to update pet spotter app model with uploaded pictures
      • During found pet, get image, call model and predict if images have a prediction score match, if match, notify owner
      • If match found, use sendgrid to send email, email asks owner to confirm if found pet is a match
    • Florence - foundational model for image, video recognition available with Azure ML
    • Responsible AI toolbox - mitigation and tracker to support look at performance of data

  • KEDA - Kubernetes Event-driven Autoscaling

    • Use to test k8s to ensure production readiness under load and autoscale

  • Scalability

    • Azure Cosmos DB set to autoscale on resource units from low to high activity with a minimum, maximum
    • Azure load testing
      • Can use open source JMeter script to simulate users and tests
    • Products
      • Azure MySQL can be auto scaled
      • Power Apps, PowerBI integration with Azure Cosmos
      • Azure Redis cache, storage of json documents
      • Virtual machine scale sets
        • VMs with baseline processing and can expand CPU

References

Section titled “References”