Columns contain attributes of record which

Provide a structure to store data in tables are inserted in rows

which consist of rows and columns You can have multiple columns

consistently writes to and reads from the

transaction-oriented applications database

OnLine Transactional Processing (OLTP) Can pull information based on say an order
Introduce
Database systems number

pull information stored on a customer by

querying Service

Stateless application

Based on the class (db.t2.micro, db.m4.large)

DB Instance hours of the DB instance consumed

Provisioned IOPS rate, regardless of IOPS

Provisioned IOPS consumed

Database storage

Charged / Pricing
Backup storage

Total of number of storage I/O requests you

I/O Requests have

internet data transfer in and out of DB

Data transfer instance

an isolated database environment running in

cloud

MySQL No limit

PostgreSQL No limit

MariaDB No limit

1 database per instance

Oracle
License included
License
Bring-Your-Own-License (BYOL)
Types
SQL Server 30 database per instance

Enterprise-grade commercial database that is

Introduce MySQL and PostgreSQL compatible

No limit

5 times better performance with MySQL

DB Instance
3 times better performance with PostgreSQL
Aurora
Main instance support read and write
workloads
Any modification data happens at primary
instance
Primary instances Each cluster has one primary

Support only read operations

Two type instances Each DB Cluster can have 15 replicas
Enable distribution of read-workloads among
instances and increase performance
Aurora Replica Can placed across AZs

5GB to 3 TB, depending on the DB instance

Magnetic engine

5GB to 6 TB for MySQL, PostgreSQL, MariaDB

and Oracle

General Purpose (SSD) 20GB to 4TB for SQL Server

Excellent for small to medium-size databases

Storage types

optimized EBS volumes

with PIOPS
optimized configuration stacks

100GB to 6TB for MySQL, MariaDB,

PostgreSQL, Oracle
Provisioned IOPS
100GB to 4TB for SQL Server Express and
Web Editions

200GB to 4TB for SQL Server Standard and

Enterprise

A cluster having two or more servers share

DB Cluster the same storage

updates to the DB instance

AWS RDS performs maintenance periodically
Update DB cluster's operating system

Most maintenance events complete in 30

minutes

Take your DB instance or DB Cluster offline

Perform maintenance on the standby

Maintenance Window Multi AZ-Deployment can further reduce the

Promote the standby to primary
impact of a maintenance event Steps

Perform maintenance on the old primary

which becomes the new standby

if you enter a maintenance value that

overlaps the backup window, an error
message appears

RDS will take the default time based on RDS

Cannot skip region if the user doesn't configure it

Increase the availability of your databases

using replication Create database clusters across multiple AZs.

Support all. Deployment for Oracle,

PostgreSQL, MariaDB, MySQL use Amazon
Technology while SQL server use SQL Server
Mirroring

Amazon automatically creates and maintains

a synchronously standby replica in a Different
AZs.

Primary instance is synchronously replicated

across AZs to standby replica to provide data
redundancy, minimise latency spikes during
system backups

High Availability

Can FORCE a failover from one AZ to another AWS Management Console

Failover by rebooting RDS Can be done
Advantages Using RebootDBInstance API call

Backup and restore are taken from

secondary which avoids I/O suspension to
the primary

RDS chooses automatically zones which is

different from master DB. You CANNOT
choose
Multi-AZ
Change CNAME to switch master and standby
DB in case of a failover

Planned database maintenance or Software

patching

Rebooting of primary instance or scaling the

RDS will automatically failover without user system
intervention when
Primary instance connectivity OR host failure
OR AZ failure
Characteristics
RDS
Loss of network connectivity to primary

Standby DB will automatically provisioned in

a different AZ with the same Region as
Primary

Synchronous connections between Primary

and Standby

Long running queries

DO NOT failover automatically in response to
database operations such as Deadlocks

Database corruption errors

Anti-patterns Not A scaling solution for read-only scenarios

Can create a second-tier Read Replica from

Aurora Max 15 Read replicas existing first-tier read replica

Can create a second-tier Read Replica from

MySQL Max 5 RR existing first-tier read replica

Support
Can create a second-tier Read Replica from
MariaDB Max 5 RR existing first-tier read replica

not currently supported to create a second-

PostgreSQL Max 5 RR tier RR

Performance Workload sharing

Benefit Durability

Availability

be read/write by setting the read_only

Read-Only connections by default parameter to false
Characteristics
Asynchronously only

Handle read traffic while source DB is in

maintenance mode

Use cases
Read replica Reporting or data warehouse tasks against
replicas

Used for scaling

Must Have automatic backups turned ON and

snapshot

Can have Read replicas of Read replicas

Key points
Can be promoted to its own database

BUT Read Replica themselves cannot be multi-

Can be built off Multi-AZ's databases AZ currently

Metric to look for is REPLICA LAG

Promote the replica to a master and keep

Disaster Recovery business work

TO support read queries across various

Cross Region Read Replicas Scaling geographic locations

For migrate DB instance quickly to another

Cross-Region migration Region

No additional charge up to 10% of the total

provisioned database storage

Charge Ex: the instance size is 20 GB. The user has

taken 2 snapshots of that DB. RDS will allow
free snapshot/automated backup up to 20GB.

Create a final snapshot and delete other

snapshots when you delete instance

transaction logs are applied to the most

appropriate daily backup in order to
restore the DB instance to the specific
Enable point-in-time recovery of DB instance When a point-in-time is initiated requested time

Full daily backup and captures transaction

logs

Automated Backups
default is 7 days

Retention period Can set to 35 days

Range from 1 to 35 days

Enable by default

Type Are deleted when the DB instance is deleted

Backup & Recovery user-initiated and enable to backup the DB

instances in a known state as frequently as
needed
Oracle

can restore to that specific state at any time

PostgreSQL

perform DB snapshots and create as many

SQL Server times as you want
Manual Snapshots
Migrate data to and from most widely used
RedShift commercial and open-source DB Snapshots keep until you explicitly delete
them even if DB instance is deleted
Aurora
Can be created with AWS Management
MariaDB Console or using CreateDBSnapshot API call

MySQL Recommend: Backup standby DB instead of

Master DB if using Multi-AZ

Ex: Oracle to Oracle Homogeneous

Can restore point-in-time backups of your
Introduce database
Ex: Oracle to MySQL
DMS 
heterogeneous
Support When restore, default DB security group is
MySQL to Aurora applied to new DB instance. If need custom
DB security, must apply them explicitly
Source and target database MUST be on AWS
service CANNOT restore to the existing DB instance.
Have to restore to a new instance.
1. Connects to source database.
2. Reads the source data, formats the data Run DB instances in a VPC and specifically in
for consumption by the target DB. a private subnet
3. Loads the data into the target DB. HowTo

Use IAM policies to assign permissions

Data analytic Use Security Groups to control what IP

Columnar data storage Data warehouse service

Aws Database address or EC2 can connect to DB instance

Security
Use SSL to encryption during transfer of data
Business Intelligence
Use RDS encryption to secure RDS and
is composed of a leader node and one or snapshots at rest
more compute nodes
Create a new DB instance with encryption
Encrypting an existing DB not support enabled and migrate data to it

By Events (RDS itself)

Cluster
BinLogDiskUsage

RedShift CPUUtilization

DatabaseConnections

DiskQueueDepth
Monitoring
Enable automated backups by default
FreeableMemory
Replicates all the data within your data
warehouse cluster when it is loaded and also FreeStorageSpace
continously backs up the data to S3

ReplicaLag (Seconds)
Retains backups for 1 day which can be
extended to max 35 days Automated backups
SwapUsage
CloudWatch by Metrics
Only backups data that has changed and are
incremental so most snapshots use up to ReadIOPS
small amount of storage
WriteIOPS
Allows manual snapshots of the data
warehouse ReadLatency

WriteLatency

ReadThroughput

WriteThroughput

NetworkReceiveThroughput

NetworkTransmitThroughput

Non-Relational Database Document-oriented database

Easy to scale horizontally for more easily

Support clustering across multiple machines

for performance and fault tolerance
Advantages
Introduce Provides both Eventually-consistent reads (by
default), strongly-consistent reads (optional)

Can NOT select a specific Availability Zone

Stateless application

built-in fault tolerance that automatically and

synchronously replicates data across three
AZs in a region
Durability & Availability
low-latency

Provisioned throughput capacity (per hour)

DynamoDB
Charged / Pricing Model Indexed data storage (per GB per month)

Data Transfer In/Out (per GB per month)

Structured data with Join and/or Complex

Transactions

Anti-patterns Large BLOB data

Large objects with low I/O rate

Store metadata for S3

Application with more oriented toward

Use cases indexing and querying data

Need pushbutton scaling or fully-automated

scaling

In-memory data store

Cache in the cloud

Retrieve information from fast, managed, in-

Introduce
Advantages Improves the performance of web apps memory data stores

Stateless application

Open-source in-memory data store and

cache, very fast

Multi-AZ

One Thread

Backup & Restore

Set

List

Data types String

Redis

Hash

Sorted set

Create Read replicas

Horizontal Scale 5 replicas max

can choose AZ

Need sorted lists, counters & hashes

Types
Use cases
Demanding applications and can scale from
single-node up to 15 shared clusters

Memory object caching system

ElastiCache
Multi-threaded performance

No Multi-AZ

No Read Replica
Caching Engines
Memcached
No backup & restore

Add cache nodes

Horizontal scale
20 nodes Max

Simple data model and you need to run large

Use cases nodes with multiple cores or threads

Open source
Common features
Master/Slave Replication

Social networking

Gaming

Use cases AdTech

Media Sharing

Q&A portals

Support for Redis cluster only

Create a backup for the cluster on a daily

basis

Automated Backups Should be performed during least bust part

Snapshot will degrade performance of the day

Retained for backup retention limit defined Max: 35 days

Allows manual snapshots of the cluster