Introduction to AWS

Session Overview

  • What is the Cloud?
  • What is AWS?
  • How to Think about Cloud Services?
  • AWS Services
    • S3
    • EC2
    • Sagemaker
    • Lambda/Fargate
    • IAM

The Cloud

The Cloud

  • Managed
  • Distributed
  • Offsite
  • On-demand
  • Scalable

Cloud Principles



Undifferentiated Heavy Lifting

AWS Regions

AWS Global Infrastructure Project Kuiper

AWS Infrastructure

AWS Data Centers

AWS Heirarchy

Why AWS?

How to Think About Cloud Services?

Typical Workflow

Mapping Workflow to the Cloud

AWS Services

AWS Service Overview

  • S3
  • EC2
  • Sagemaker
  • Lambda/Fargate
  • IAM

Simple Storage Service (S3)

One of the modern wonders of the world. Planet Scale Storage System

  • Object Storage
  • 99.999999999% Durability
  • 99.99% Availability
  • The base for everything!
  • Organized by Globally Unique Buckets
  • Multiple Tiers (Including long term storage in Glacier)

Object Storage

  • Not a file system.
  • Flat - not heirarchical.
  • Data is an object.
  • Metadata!
  • Stored under keys.
  • Massively scalable across structured and unstructured data with metadata!
  • REST access

Performance

The Foundation for Performant Data Science:

Object Storage = Performance

  • Massively Parallel
  • Massively Scalable: Serverless
    • Supported Storage
    • Supported Access
  • Massively Reliable :)

Using AWS S3

REST API:

  • GET, PUT, HEAD

AWS CLI:

  • aws s3 cp
  • aws s3 sync
  • aws s3 ls

Python

AWS Services

AWS Service Overview

  • S3
  • EC2
  • Sagemaker
  • Lambda/Fargate
  • IAM

Elastic Compute Cloud (EC2)

  • Massively Scalable Compute
  • Massively Configurable
  • 99.99% Availability
  • The base for everything!
  • On-demand, Reserved instances
  • Elasticity can be automated
  • Attachable storage
  • First step toward managed compute

EC2 Notes

  • Your flexible computer in the cloud and so much more.
  • Networking and security are extremely important.
  • Compute and storage are independent and ephemeral.
  • Coder is running templated EC2 instances

Sagemaker

  • Special, managed EC2
  • Configured for Machine Learning
  • ML Ecosystem
  • Preconfigured Networking with Reasonable Defaults

Lambda/Fargate

  • Managed Compute
  • Serverless
  • Lambda - Function as a Service
  • Fargate - Container as a Service
  • Easily burst scalable

AWS Services

AWS Service Overview

  • S3
  • EC2
  • Sagemaker
  • Lambda/Fargate
  • IAM

Identity Access Management

  • Users
  • Roles
  • Policies

** Access Denied or Unauthorized **

** Mostly Managed through SSO **

AWS Services

AWS Building Blocks