Overall

20% more expensive than rds.

Copy 6 copies of user data across 3 AZs

• 4 copies of 6 need for write

• 3 copies of 6 need for read

One aurora instance takes write master

Automated fail over is less than 3 seconds

Master + up to 15 aurora replica serve reads

Support for cross region replications

Aurora autoscaling can scale up to 15 read replica across different AZ

Writer endpoint and reader endpoint

Can create custom endpoint for a certain subset of read replica

Reader endpoint generally not used after read replica

Aurora Serverless

Pay per second

Client hit the proxy fleet (managed by aurora)

Global Aurora

Aurora Cross Region read replica

• useful for disaster recover

• simple to put in place

Aurora Global Database

• 1 primary region(read/write)

• Up to 10 secondary regions(read replica), replication lag is less than 1 second

• Up to 16 read replica for secondary regions

• Cross region replication is less than 1 second

Aurora Machine learning

Can add ML prediction to ur app via sql

Optimized integration with AWS ML Services

Supported services

• Sage maker

• Comprehend

Fraud detection, ads targeting, sentiment analysis, product recommendations

Babelfish

Allow aurora Postgres to work with MSQL commands

Backups

Automated backups

1 to 35 days can’t be disabled

Point in time recovery at anytime

Manual DB snapshot

Manually triggered by the user

Retention backup for as long as you want

Restore

Create backup of on-premise database using persona xtrabackup

Store file in s3

Restore backup on aurora instance running mysql

Clone

Faster than snapshot and restore

Fast and cost effective

Used for cloning to stg env

Uses copy on write protocol

(After cloning, if new db get updated, the db will be separated)

(If new cluster haven’t updated yet, the original db update will reflected the cloned one)