CMSC389L

Lambda

Lambda: What is it?

AWS Lambda lets you run code without provisioning or managing servers. You pay only for the compute time you consume - there is no charge when your code is not running. With Lambda, you can run code for virtually any type of application or backend service - all with zero administration. Just upload your code and Lambda takes care of everything required to run and scale your code with high availability.

Serverless: Historical Context

• Data Centers: (pre-AWS)
  • Maintain hardware! 😷
• Infrastructure as a Service: EC2
  • Configure, manage, & scale servers
• Platform as a Service: Beanstalk
  • Configure, manage OS/patches/etc.
• Containers: Docker, ECS
  • Have to scale # of containers
• Serverless: Lambda
  • Just write code -- no operational overhead

Serverless on AWS

• Three key components of application architectures

Lambda: Benefits

• Zero administration: you don't need to manage, patch, and scale servers
• Continuous scaling: Lambda functions naturally scale horizontally, and you don't need to provision more instances as your traffic goes up!
• Subsecond metering: Metering down to 100ms, which is 10x better than EC2
  • You only pay for your usage -- you don't pay for idle time
• Security: Principle of Least Privilege: can configure IAM policies per-function, rather than per-application/per-instance
• Microservice Architecture: FaaS naturally encourages this

Lambda: What does it look like?

Lambda: Complexity

• Monolith vs Services/Functions:
  • Monolith services are easy to reason about at the architecture level.
  • Microservices / Functions lead to complicated relationships between functions.
• Why is that okay?
  • Forces you to logically separate components.
    • Therefore, easy to add/remove features (good for reducing tech debt).
  • Much easier to develop on with a team, reduces organizational overhead.

Lambda: Dealing w/ Complexity

• Can be tricky to monitor/debug via CloudWatch logs.
  • Use AWS X-Ray for distributed tracing, track requests between functions.

Lambda: Concepts

• Function: The code that you are executing. Supported languages: Node, Python, Java, C#.
• Handler: The main function in your code.
• Template: Amazon provides a number of templates that you can use to launch basic Lambda functions
• Role: IAM role that the function executes under. Necessary for security and debuggability.
• Deployment Package: ZIP directory containing all code/dependencies.
  • Can include your dependencies or other necessary assets

Lambda: Concepts (cont.)

• Trigger: An event that invokes a Lambda function.
• Timeout: Lambda functions are limited to 5m, but you should set this timeout lower depending on 
• Sync vs. Async: Synchronous mode will return a value, async just fires off the function.
  • Async mode will receive an immediate response, indicating if the function started correctly. 

Lambda: Triggers

• Lambda functions are invoked by some kind of trigger (aka event). For example:
  • S3 (on upload, metadata change, removal, ...)
  • DynamoDB (insert, update, delete, ...)
  • CodeCommit (on git commit, ...) 
  • Scheduled (CRON-style), via CloudWatch Events
  • Directly (over HTTP, say from a client's browser) via API Gateway
• In other words, Lambda enables event-based programming.

Lambda: Example Architecture

Lambda: How do they work?

• Similar to how EC2 uses a hypervisor to isolate instances, Lambda uses Linux containers (LXC).
  • Containers are extremely lightweight -- no OS duplication.
  • Amazon will run 100s of these on a single physical server.
• Controlled environment (libraries, language, etc.) means that AWS can eek out the efficiency of these instances.

Lambda @ Edge

• CDN's are great because they bring static content as close to your users as possible
  • Reduce latencies, etc.
• What if you could do that with compute?
• L@E: Lets you run Lambda requests in between server/CloudFront or user/CloudFront
  • Can inspect HTTP headers, perform logic such as redirection

Lambda Pricing

• Two pricing components: # requests, and duration

• # Requests
  • $0.20 / 1M requests
  • Free Tier: first 1M requests

• Duration
  • Pay per ms of execution time, rounded up to 100ms
  • $16.64 / 1M GB-second
    • = $0.208 / 1M 128MB-100ms (min RAM & shortest time)
  • Free Tier: 400,000 GB-seconds
    • you could run a 128MB Lambda function continuously for an entire month under the free tier

Lambda: Optimizations

• Containers can be re-used. Move as much re-useable logic outside of the handler. Global state will be preserved.
  • For example: database connections; avoid variable re-initialization
• Minimize the size of your deployment package (code, etc.)
  • Prefer small frameworks and libraries
• Test your Lambda code on various memory configurations
  • May reduce your overall costs, depending on your task
• Load test your Lambda function to determine                        the minimum necessary timeout value
  • Lambda functions that fail will terminate                    sooner, saving on duration costs

Lambda Demo

Lambda Worksheet

Complete the worksheet section on Lambda.

Final Project Details

(Get pumped)

Final Project

• Goal: Build a meaningful project that proves your AWS skillset to employers.
  • Develop experience interconnecting different AWS services to build a project in a scalable fashion.
• Requirements:
  • Must use at least 3 different AWS services
  • At least one of which is a service we have not covered in class
    • SaaS: Polly (text to speech)
    • IaaS: RDS (MySQL, PostgreSQL, etc.)
    • Batch: AWS MapReduce (big data processing)
    • Other: AWS IoT (programmable Dash buttons!)

Deliverables

• Proposal
  • Formatted as slides. Submit online, demo in-person to a facilitator.
  • Due this Sunday
• Checkpoints
  • Status update on progress, roadblocks
  • First Checkpoint: by 4/13
• Code + Demos
  • Public GitHub repo
  • 3-5m YouTube demo video (I recommend using Loom)
  • In-person demo with a facilitator

Final Project: Examples

• Alexa App
  • Build an Alexa app that can play you in chess.
  • Use AWS Alexa Skill Kit and Lambda to respond to user requests.
  • Store user data and game data in DynamoDB.
  • Optionally, build a front-end for it on top of Elastic Beanstalk or S3!
• Some other options:
  • AWS IoT: You can access hardware for free through Sandbox!
  • Chatbot: Check out AWS Lex!
  • etc.