Ref: https://learn.cantrill.io/courses/1820301/lectures/41301398

• Summary:
• Contents:

Serverless - Basic Concepts

• 🔧 Main idea: manage few (if any) servers
  • 💡 Serverless isn't one single thing… it's a concept more than anything
    • More a SW architecture than a HW architecture!
  • Servers carry overhead: cost, administration & risk
    • 👍 Without server management, overhead is greatly reduced
  • Serverless architecture takes the best things from microservices & EDA
• ‼️ Despite the name “serverless” there ARE actually servers running behind the scenes!! xD
  • …You just don't manage them!
• AWS has tons of managed services and products that support serverless architectures
  • 💡 Don't expect to know them all. More important is to understand how serverless works.

Characteristics of Serverless Applications

• Apps are a collection of small & specialized functions
  • Even smaller than servers in microservices!
  • Functions do small things very, very well. They start, execute & stop.
  • Functions only bill for resources consumed during their execution
• Stateless and ephemeral environments
  • Functions can run anywhere
• Event-driven
  • Consumption only when being used → consumption-based model
  • Little, if any, costs incurred for base usage (i.e. no events & nothing running)
• Managed services used where possible → don't reinvent the wheel
  • e.g. S3 for persistent object storage, DynamoDB for persistent data storage, External ID providers instead of building your own…
  • ❗ Consume as a service as much as possible, code as little as possible
• FaaS used where possible for general purpose compute, if needed
  • Cheap & easy to scale. Each environment is easy to provision & predictably the same
  • In AWS: Lambda (in Azure: Azure Functions)
  • Default is to NOT use self-managed compute (i.e. no servers, no EC2 instances)

Example of a Simple Serverless Architecture

Diagram: https://github.com/acantril/aws-sa-associate-saac03/blob/main/1600-SERVERLESS_and_APPLICATION_SERVICES/00_LEARNINGAIDS/Serverless-1.png

• 💡 We continue with CatTube example app from previous Architecture Deep Dive section
  • 💡 Simplified architecture → in reality we would have e.g. APIGW between client and Lambda functions
  • ‼️ Notice how there are absolutely NO managed servers in the whole architecture!!!
    • No self-managed DB servers either!!
• Frontend: S3 static website with HTML pages that inject JavaScript to browser
  • Modern browsers can run client-side JavaScript inside the browser
• Authentication: Third-party identity provider (e.g. Google IDP)
  1. IDP returns ID token after successful authentication
  2. AWS can't use third-party ID tokens → browser communicates with Amazon Cognito
  3. Cognito swaps IDP token for AWS temporary credentials (with an IAM role) → client can now access AWS resources