Aws Series High Availability and Scaling

Aws Series High Availability and Scaling

Post Date : 2023-12-30T09:31:02+07:00

Modified Date : 2023-12-30T09:31:02+07:00

Category: systemdesign aws

Tags: aws

Vertical Scaling vs Horizontal Scaling

Vertical Scaling


  • Increase your EC2 RAM
  • Increase your EC2 CPU


Horizontal Scaling


  • Add new EC2
  • Add new EC2


But we dont’ only scale EC2, it should include other parts:

  • Database
  • VPC
  • AZ

The 3W’s of Scaling

What do we scale?

  • What sort of resource, we’re going to scale. How do we define template?

Where do we scale?

  • When applying the model, where does it go? Should we scale out database? Or Webservers?

When do we scale?

  • How do we know we need more? CloudWatch alarms can tell us it’s time to add more resources.

Launch Template

  • A launch template specifies the needed settings that go into building out an EC2 instance. It is a collection of settings you can configure so you dont’ have to walk through the EC2 wizard over and over.

It’s a heart of auto scaling feature.

Let’s compare Lauch Template with Configurations:

Launch Template

  • Capable of leveraging all EC2 auto scaling features.
  • Supports versioning
  • More granularity
  • AWS recommends this approach


  • Only for certain EC2 auto scaling features
  • Immutable
  • Limited configuration options
  • Try not to use them

What can be included inside a Launch Template?

  • AMI
  • EC Instance Size
  • Security Groups
  • Potentially networking information( if you use autoscaling, this part should be set inside autoscaling group configuration)
  • Potentially UserData

Scaling EC2 Instances with Auto Scaling Groups

  • Creating a High Available application
  • Remember spread the resources out over multiple AZs and utilize load balancers

Auto Scaling Groups

  • Contain a collection of EC2 Instances that are treated as collective group for purposes of scaling and management.

Auto Scaling Steps




Step Scaling




Simple Scaling

  • Relies on metrics for scaling needs

Example: Add 1 instances if CPU Utilization metric > 80%

Target Tracking

  • Uses a Scaling metric and value that ASG should maintain at all times

Eg: Maintain ASGAverageCPUUtilization = 50%

Instance Warm-up and CoolDown

Warm-Up: stops instances from being placed behind the load balancer, failing the health check, and being terminated prematurely

CoolDown: Pauses Auto Scaling for a set amount of time. Help you avoid runaway scaling events. -> prevent over doing auto scaling

Avoid Thrashing: You want to create instances quickly and spin them down slowly

Scaling Types

  • Reactive Scaling: Measure and determine if you need to create/remove resources
  • Schedule Scaling: Have a predictable workload, create a scaling event to get your resources ready to go before they’re actually needed -> shopping events(black Friday, Christmas, Hoildays)
  • Predictive Scaling: AWS uses it machine learning algorithms to determine when you’ll need to scale. They are reevaluated every 24 hours to create a forecast for the next 48 hours

Key Figures

  • Desired: initial value
  • Min: minimum value
  • Max: maximum value

Scaling Relational Databases

4 Types of Scaling


  • Vertical Scaling : resize
  • Scaling Storage: storage can be resized, but it’s only able to go up, not down
  • Read replicas: creating read-only copies of your data can help spread out the workload
  • Aurora serverless: We can offload autoscale to AWS. Excels with unpredictable workloads

Scaling Non-Relational Databases


  • AWS does all the heavy lifting for you
  • Just 2 options: provisioned(predictable workload) and on-demand(sporadic workload)
  • You can switch between 2 options, but only twice per 24hours per table


Some figures

RCU(read capacity unit)

  • DynamoDB Unit of measurement for reads per second for an item up to 4KB in size.
  • One strongly consistent read per second
  • Two eventually consistent read per second

So how many RCUs for 1 strongly consistent read per second for object that are 7KB in size?

1 RCU = 4 KB/ 1 strongly consistent read Round up to the next nearest amount for the item size = 8KB/4KB = 2 RCU

WCU(write capacity unit)

  • DynanoDB Unit of measurement for writes per second for an item up to 1KB in size

So how many WCUs for 1 write per second for an object that is 3KB in size?

1 WCU = 1KB * 1 write per second
3KB * 1 WCU = 3 WCU

Disaster Recovery Strategies

RPO (Recovery Point Objective)

  • How much data you can afford to lose?
  • Typically speaking, the lower time, the greater the cost

RTO (Recovery Time Objective)

  • How fast do you want to fail over? How much time can the business afford?
  • Typically speaking, the lower the time, the more expensive the cost

Backup and Restore

Simplest Disaster Recovery Strategy

  • Restore the system from your backup

Pilot Light

image image image

Warm StandBy

image image image

Active/Active Failover

image image