AI

Workflow of a Machine Learning Project .

Example : Amazon Echo
1.collect data
2.Train Model : Iterate many times until good enough .
3. Deploy Model : get data back , maintain update model .

Workflow of a Data Science Project .
1.Collect Data
2. Analyze Data : Iterate many times to get good insight
3. Suggest hypothesis /action : Deploy changes / reanalyze new Data periodically .

Aws ElasticCache

Web application
Database type : radis

Aws database services

Aws database migration service

Redshirt data warehouse solution.

Petabyte scale .

Wide column storage
Graph database.
Neptune

RDS database snapshot is placed to S3.
This is an increment backup.

Aurora : reduce cost by 90%
Dynamo db runs on top of s3

Aws CloudFront

CDN

User will go to edge location rather than origin.

Amazon S3

Simple Storage Service

• Files can be stored and retrieved from  anywhere on the web .

• Used for live streaming . Netflix, Hotstar use it for hosting.

• Used for content distribution , big data analytics , backup etc .

• Files are stored in buckets , a max of 100 buckets can be created by a developer .
  
• Replica of the file is available in 3 different Availability Zones , all in the same region . AWS also supports cross region replication which can be configured manually .
  
• Supports periodic automatic archival/deletion , using lifecycle rules.
  

Storage class
Standard available in 3 availability zone
Glacier : cold data.
Archival data ..
Glacier deep
Long data retention

Requester pays
The one who download pays

Object lifecycle management
Move file automatically to glacier after 50 days
Cannot move back from glacier to S3 via lifecycle

Retreival options

EFS : multiple EC2 share same storage .
Nfs : network file system.

Snowball.
Hardware device to migrate a data .
Like a briefcase .
Device will be shipped.

Aws elastic block storage

IaaS

Infrastructure as a Service  :

 provides access to virtual or physical hardware .

EBS

     Elastic Block Storage

• Similar to a hard drive
• Block storage for EC2 and other services
• Automatic replication within its AZ
• Snapshot backup to Amazon S3
• Restore a snapshot to a new EBS volume
• Multiple volumes can be attached a single EC2
  
• Volumes can be shared between multiple EC2
  

SSD
HDD

Aws lambda

Serverless

Function as a service
I want to shut down my services at 6 pm.
Integrated with Database.

Build in fault tolerance.
Pricing: only when it is executed.

Create thumbnail function